sqlglot.parser

   1from __future__ import annotations
   2
   3import logging
   4import typing as t
   5import itertools
   6from collections import defaultdict
   7
   8from sqlglot import exp
   9from sqlglot.errors import ErrorLevel, ParseError, concat_messages, merge_errors
  10from sqlglot.helper import apply_index_offset, ensure_list, seq_get
  11from sqlglot.time import format_time
  12from sqlglot.tokens import Token, Tokenizer, TokenType
  13from sqlglot.trie import TrieResult, in_trie, new_trie
  14
  15if t.TYPE_CHECKING:
  16    from sqlglot._typing import E, Lit
  17    from sqlglot.dialects.dialect import Dialect, DialectType
  18
  19    T = t.TypeVar("T")
  20    TCeilFloor = t.TypeVar("TCeilFloor", exp.Ceil, exp.Floor)
  21
  22logger = logging.getLogger("sqlglot")
  23
  24OPTIONS_TYPE = t.Dict[str, t.Sequence[t.Union[t.Sequence[str], str]]]
  25
  26
  27def build_var_map(args: t.List) -> exp.StarMap | exp.VarMap:
  28    if len(args) == 1 and args[0].is_star:
  29        return exp.StarMap(this=args[0])
  30
  31    keys = []
  32    values = []
  33    for i in range(0, len(args), 2):
  34        keys.append(args[i])
  35        values.append(args[i + 1])
  36
  37    return exp.VarMap(keys=exp.array(*keys, copy=False), values=exp.array(*values, copy=False))
  38
  39
  40def build_like(args: t.List) -> exp.Escape | exp.Like:
  41    like = exp.Like(this=seq_get(args, 1), expression=seq_get(args, 0))
  42    return exp.Escape(this=like, expression=seq_get(args, 2)) if len(args) > 2 else like
  43
  44
  45def binary_range_parser(
  46    expr_type: t.Type[exp.Expression], reverse_args: bool = False
  47) -> t.Callable[[Parser, t.Optional[exp.Expression]], t.Optional[exp.Expression]]:
  48    def _parse_binary_range(
  49        self: Parser, this: t.Optional[exp.Expression]
  50    ) -> t.Optional[exp.Expression]:
  51        expression = self._parse_bitwise()
  52        if reverse_args:
  53            this, expression = expression, this
  54        return self._parse_escape(self.expression(expr_type, this=this, expression=expression))
  55
  56    return _parse_binary_range
  57
  58
  59def build_logarithm(args: t.List, dialect: Dialect) -> exp.Func:
  60    # Default argument order is base, expression
  61    this = seq_get(args, 0)
  62    expression = seq_get(args, 1)
  63
  64    if expression:
  65        if not dialect.LOG_BASE_FIRST:
  66            this, expression = expression, this
  67        return exp.Log(this=this, expression=expression)
  68
  69    return (exp.Ln if dialect.parser_class.LOG_DEFAULTS_TO_LN else exp.Log)(this=this)
  70
  71
  72def build_hex(args: t.List, dialect: Dialect) -> exp.Hex | exp.LowerHex:
  73    arg = seq_get(args, 0)
  74    return exp.LowerHex(this=arg) if dialect.HEX_LOWERCASE else exp.Hex(this=arg)
  75
  76
  77def build_lower(args: t.List) -> exp.Lower | exp.Hex:
  78    # LOWER(HEX(..)) can be simplified to LowerHex to simplify its transpilation
  79    arg = seq_get(args, 0)
  80    return exp.LowerHex(this=arg.this) if isinstance(arg, exp.Hex) else exp.Lower(this=arg)
  81
  82
  83def build_upper(args: t.List) -> exp.Upper | exp.Hex:
  84    # UPPER(HEX(..)) can be simplified to Hex to simplify its transpilation
  85    arg = seq_get(args, 0)
  86    return exp.Hex(this=arg.this) if isinstance(arg, exp.Hex) else exp.Upper(this=arg)
  87
  88
  89def build_extract_json_with_path(expr_type: t.Type[E]) -> t.Callable[[t.List, Dialect], E]:
  90    def _builder(args: t.List, dialect: Dialect) -> E:
  91        expression = expr_type(
  92            this=seq_get(args, 0), expression=dialect.to_json_path(seq_get(args, 1))
  93        )
  94        if len(args) > 2 and expr_type is exp.JSONExtract:
  95            expression.set("expressions", args[2:])
  96
  97        return expression
  98
  99    return _builder
 100
 101
 102def build_mod(args: t.List) -> exp.Mod:
 103    this = seq_get(args, 0)
 104    expression = seq_get(args, 1)
 105
 106    # Wrap the operands if they are binary nodes, e.g. MOD(a + 1, 7) -> (a + 1) % 7
 107    this = exp.Paren(this=this) if isinstance(this, exp.Binary) else this
 108    expression = exp.Paren(this=expression) if isinstance(expression, exp.Binary) else expression
 109
 110    return exp.Mod(this=this, expression=expression)
 111
 112
 113def build_pad(args: t.List, is_left: bool = True):
 114    return exp.Pad(
 115        this=seq_get(args, 0),
 116        expression=seq_get(args, 1),
 117        fill_pattern=seq_get(args, 2),
 118        is_left=is_left,
 119    )
 120
 121
 122def build_array_constructor(
 123    exp_class: t.Type[E], args: t.List, bracket_kind: TokenType, dialect: Dialect
 124) -> exp.Expression:
 125    array_exp = exp_class(expressions=args)
 126
 127    if exp_class == exp.Array and dialect.HAS_DISTINCT_ARRAY_CONSTRUCTORS:
 128        array_exp.set("bracket_notation", bracket_kind == TokenType.L_BRACKET)
 129
 130    return array_exp
 131
 132
 133def build_convert_timezone(
 134    args: t.List, default_source_tz: t.Optional[str] = None
 135) -> t.Union[exp.ConvertTimezone, exp.Anonymous]:
 136    if len(args) == 2:
 137        source_tz = exp.Literal.string(default_source_tz) if default_source_tz else None
 138        return exp.ConvertTimezone(
 139            source_tz=source_tz, target_tz=seq_get(args, 0), timestamp=seq_get(args, 1)
 140        )
 141
 142    return exp.ConvertTimezone.from_arg_list(args)
 143
 144
 145def build_trim(args: t.List, is_left: bool = True):
 146    return exp.Trim(
 147        this=seq_get(args, 0),
 148        expression=seq_get(args, 1),
 149        position="LEADING" if is_left else "TRAILING",
 150    )
 151
 152
 153def build_coalesce(
 154    args: t.List, is_nvl: t.Optional[bool] = None, is_null: t.Optional[bool] = None
 155) -> exp.Coalesce:
 156    return exp.Coalesce(this=seq_get(args, 0), expressions=args[1:], is_nvl=is_nvl, is_null=is_null)
 157
 158
 159def build_locate_strposition(args: t.List):
 160    return exp.StrPosition(
 161        this=seq_get(args, 1),
 162        substr=seq_get(args, 0),
 163        position=seq_get(args, 2),
 164    )
 165
 166
 167class _Parser(type):
 168    def __new__(cls, clsname, bases, attrs):
 169        klass = super().__new__(cls, clsname, bases, attrs)
 170
 171        klass.SHOW_TRIE = new_trie(key.split(" ") for key in klass.SHOW_PARSERS)
 172        klass.SET_TRIE = new_trie(key.split(" ") for key in klass.SET_PARSERS)
 173
 174        return klass
 175
 176
 177class Parser(metaclass=_Parser):
 178    """
 179    Parser consumes a list of tokens produced by the Tokenizer and produces a parsed syntax tree.
 180
 181    Args:
 182        error_level: The desired error level.
 183            Default: ErrorLevel.IMMEDIATE
 184        error_message_context: The amount of context to capture from a query string when displaying
 185            the error message (in number of characters).
 186            Default: 100
 187        max_errors: Maximum number of error messages to include in a raised ParseError.
 188            This is only relevant if error_level is ErrorLevel.RAISE.
 189            Default: 3
 190    """
 191
 192    FUNCTIONS: t.Dict[str, t.Callable] = {
 193        **{name: func.from_arg_list for name, func in exp.FUNCTION_BY_NAME.items()},
 194        **dict.fromkeys(("COALESCE", "IFNULL", "NVL"), build_coalesce),
 195        "ARRAY": lambda args, dialect: exp.Array(expressions=args),
 196        "ARRAYAGG": lambda args, dialect: exp.ArrayAgg(
 197            this=seq_get(args, 0), nulls_excluded=dialect.ARRAY_AGG_INCLUDES_NULLS is None or None
 198        ),
 199        "ARRAY_AGG": lambda args, dialect: exp.ArrayAgg(
 200            this=seq_get(args, 0), nulls_excluded=dialect.ARRAY_AGG_INCLUDES_NULLS is None or None
 201        ),
 202        "CHAR": lambda args: exp.Chr(expressions=args),
 203        "CHR": lambda args: exp.Chr(expressions=args),
 204        "COUNT": lambda args: exp.Count(this=seq_get(args, 0), expressions=args[1:], big_int=True),
 205        "CONCAT": lambda args, dialect: exp.Concat(
 206            expressions=args,
 207            safe=not dialect.STRICT_STRING_CONCAT,
 208            coalesce=dialect.CONCAT_COALESCE,
 209        ),
 210        "CONCAT_WS": lambda args, dialect: exp.ConcatWs(
 211            expressions=args,
 212            safe=not dialect.STRICT_STRING_CONCAT,
 213            coalesce=dialect.CONCAT_COALESCE,
 214        ),
 215        "CONVERT_TIMEZONE": build_convert_timezone,
 216        "DATE_TO_DATE_STR": lambda args: exp.Cast(
 217            this=seq_get(args, 0),
 218            to=exp.DataType(this=exp.DataType.Type.TEXT),
 219        ),
 220        "GENERATE_DATE_ARRAY": lambda args: exp.GenerateDateArray(
 221            start=seq_get(args, 0),
 222            end=seq_get(args, 1),
 223            step=seq_get(args, 2) or exp.Interval(this=exp.Literal.string(1), unit=exp.var("DAY")),
 224        ),
 225        "GLOB": lambda args: exp.Glob(this=seq_get(args, 1), expression=seq_get(args, 0)),
 226        "HEX": build_hex,
 227        "JSON_EXTRACT": build_extract_json_with_path(exp.JSONExtract),
 228        "JSON_EXTRACT_SCALAR": build_extract_json_with_path(exp.JSONExtractScalar),
 229        "JSON_EXTRACT_PATH_TEXT": build_extract_json_with_path(exp.JSONExtractScalar),
 230        "LIKE": build_like,
 231        "LOG": build_logarithm,
 232        "LOG2": lambda args: exp.Log(this=exp.Literal.number(2), expression=seq_get(args, 0)),
 233        "LOG10": lambda args: exp.Log(this=exp.Literal.number(10), expression=seq_get(args, 0)),
 234        "LOWER": build_lower,
 235        "LPAD": lambda args: build_pad(args),
 236        "LEFTPAD": lambda args: build_pad(args),
 237        "LTRIM": lambda args: build_trim(args),
 238        "MOD": build_mod,
 239        "RIGHTPAD": lambda args: build_pad(args, is_left=False),
 240        "RPAD": lambda args: build_pad(args, is_left=False),
 241        "RTRIM": lambda args: build_trim(args, is_left=False),
 242        "SCOPE_RESOLUTION": lambda args: exp.ScopeResolution(expression=seq_get(args, 0))
 243        if len(args) != 2
 244        else exp.ScopeResolution(this=seq_get(args, 0), expression=seq_get(args, 1)),
 245        "STRPOS": exp.StrPosition.from_arg_list,
 246        "CHARINDEX": lambda args: build_locate_strposition(args),
 247        "INSTR": exp.StrPosition.from_arg_list,
 248        "LOCATE": lambda args: build_locate_strposition(args),
 249        "TIME_TO_TIME_STR": lambda args: exp.Cast(
 250            this=seq_get(args, 0),
 251            to=exp.DataType(this=exp.DataType.Type.TEXT),
 252        ),
 253        "TO_HEX": build_hex,
 254        "TS_OR_DS_TO_DATE_STR": lambda args: exp.Substring(
 255            this=exp.Cast(
 256                this=seq_get(args, 0),
 257                to=exp.DataType(this=exp.DataType.Type.TEXT),
 258            ),
 259            start=exp.Literal.number(1),
 260            length=exp.Literal.number(10),
 261        ),
 262        "UNNEST": lambda args: exp.Unnest(expressions=ensure_list(seq_get(args, 0))),
 263        "UPPER": build_upper,
 264        "VAR_MAP": build_var_map,
 265    }
 266
 267    NO_PAREN_FUNCTIONS = {
 268        TokenType.CURRENT_DATE: exp.CurrentDate,
 269        TokenType.CURRENT_DATETIME: exp.CurrentDate,
 270        TokenType.CURRENT_TIME: exp.CurrentTime,
 271        TokenType.CURRENT_TIMESTAMP: exp.CurrentTimestamp,
 272        TokenType.CURRENT_USER: exp.CurrentUser,
 273    }
 274
 275    STRUCT_TYPE_TOKENS = {
 276        TokenType.NESTED,
 277        TokenType.OBJECT,
 278        TokenType.STRUCT,
 279        TokenType.UNION,
 280    }
 281
 282    NESTED_TYPE_TOKENS = {
 283        TokenType.ARRAY,
 284        TokenType.LIST,
 285        TokenType.LOWCARDINALITY,
 286        TokenType.MAP,
 287        TokenType.NULLABLE,
 288        TokenType.RANGE,
 289        *STRUCT_TYPE_TOKENS,
 290    }
 291
 292    ENUM_TYPE_TOKENS = {
 293        TokenType.DYNAMIC,
 294        TokenType.ENUM,
 295        TokenType.ENUM8,
 296        TokenType.ENUM16,
 297    }
 298
 299    AGGREGATE_TYPE_TOKENS = {
 300        TokenType.AGGREGATEFUNCTION,
 301        TokenType.SIMPLEAGGREGATEFUNCTION,
 302    }
 303
 304    TYPE_TOKENS = {
 305        TokenType.BIT,
 306        TokenType.BOOLEAN,
 307        TokenType.TINYINT,
 308        TokenType.UTINYINT,
 309        TokenType.SMALLINT,
 310        TokenType.USMALLINT,
 311        TokenType.INT,
 312        TokenType.UINT,
 313        TokenType.BIGINT,
 314        TokenType.UBIGINT,
 315        TokenType.INT128,
 316        TokenType.UINT128,
 317        TokenType.INT256,
 318        TokenType.UINT256,
 319        TokenType.MEDIUMINT,
 320        TokenType.UMEDIUMINT,
 321        TokenType.FIXEDSTRING,
 322        TokenType.FLOAT,
 323        TokenType.DOUBLE,
 324        TokenType.UDOUBLE,
 325        TokenType.CHAR,
 326        TokenType.NCHAR,
 327        TokenType.VARCHAR,
 328        TokenType.NVARCHAR,
 329        TokenType.BPCHAR,
 330        TokenType.TEXT,
 331        TokenType.MEDIUMTEXT,
 332        TokenType.LONGTEXT,
 333        TokenType.BLOB,
 334        TokenType.MEDIUMBLOB,
 335        TokenType.LONGBLOB,
 336        TokenType.BINARY,
 337        TokenType.VARBINARY,
 338        TokenType.JSON,
 339        TokenType.JSONB,
 340        TokenType.INTERVAL,
 341        TokenType.TINYBLOB,
 342        TokenType.TINYTEXT,
 343        TokenType.TIME,
 344        TokenType.TIMETZ,
 345        TokenType.TIMESTAMP,
 346        TokenType.TIMESTAMP_S,
 347        TokenType.TIMESTAMP_MS,
 348        TokenType.TIMESTAMP_NS,
 349        TokenType.TIMESTAMPTZ,
 350        TokenType.TIMESTAMPLTZ,
 351        TokenType.TIMESTAMPNTZ,
 352        TokenType.DATETIME,
 353        TokenType.DATETIME2,
 354        TokenType.DATETIME64,
 355        TokenType.SMALLDATETIME,
 356        TokenType.DATE,
 357        TokenType.DATE32,
 358        TokenType.INT4RANGE,
 359        TokenType.INT4MULTIRANGE,
 360        TokenType.INT8RANGE,
 361        TokenType.INT8MULTIRANGE,
 362        TokenType.NUMRANGE,
 363        TokenType.NUMMULTIRANGE,
 364        TokenType.TSRANGE,
 365        TokenType.TSMULTIRANGE,
 366        TokenType.TSTZRANGE,
 367        TokenType.TSTZMULTIRANGE,
 368        TokenType.DATERANGE,
 369        TokenType.DATEMULTIRANGE,
 370        TokenType.DECIMAL,
 371        TokenType.DECIMAL32,
 372        TokenType.DECIMAL64,
 373        TokenType.DECIMAL128,
 374        TokenType.DECIMAL256,
 375        TokenType.UDECIMAL,
 376        TokenType.BIGDECIMAL,
 377        TokenType.UUID,
 378        TokenType.GEOGRAPHY,
 379        TokenType.GEOMETRY,
 380        TokenType.POINT,
 381        TokenType.RING,
 382        TokenType.LINESTRING,
 383        TokenType.MULTILINESTRING,
 384        TokenType.POLYGON,
 385        TokenType.MULTIPOLYGON,
 386        TokenType.HLLSKETCH,
 387        TokenType.HSTORE,
 388        TokenType.PSEUDO_TYPE,
 389        TokenType.SUPER,
 390        TokenType.SERIAL,
 391        TokenType.SMALLSERIAL,
 392        TokenType.BIGSERIAL,
 393        TokenType.XML,
 394        TokenType.YEAR,
 395        TokenType.USERDEFINED,
 396        TokenType.MONEY,
 397        TokenType.SMALLMONEY,
 398        TokenType.ROWVERSION,
 399        TokenType.IMAGE,
 400        TokenType.VARIANT,
 401        TokenType.VECTOR,
 402        TokenType.VOID,
 403        TokenType.OBJECT,
 404        TokenType.OBJECT_IDENTIFIER,
 405        TokenType.INET,
 406        TokenType.IPADDRESS,
 407        TokenType.IPPREFIX,
 408        TokenType.IPV4,
 409        TokenType.IPV6,
 410        TokenType.UNKNOWN,
 411        TokenType.NOTHING,
 412        TokenType.NULL,
 413        TokenType.NAME,
 414        TokenType.TDIGEST,
 415        TokenType.DYNAMIC,
 416        *ENUM_TYPE_TOKENS,
 417        *NESTED_TYPE_TOKENS,
 418        *AGGREGATE_TYPE_TOKENS,
 419    }
 420
 421    SIGNED_TO_UNSIGNED_TYPE_TOKEN = {
 422        TokenType.BIGINT: TokenType.UBIGINT,
 423        TokenType.INT: TokenType.UINT,
 424        TokenType.MEDIUMINT: TokenType.UMEDIUMINT,
 425        TokenType.SMALLINT: TokenType.USMALLINT,
 426        TokenType.TINYINT: TokenType.UTINYINT,
 427        TokenType.DECIMAL: TokenType.UDECIMAL,
 428        TokenType.DOUBLE: TokenType.UDOUBLE,
 429    }
 430
 431    SUBQUERY_PREDICATES = {
 432        TokenType.ANY: exp.Any,
 433        TokenType.ALL: exp.All,
 434        TokenType.EXISTS: exp.Exists,
 435        TokenType.SOME: exp.Any,
 436    }
 437
 438    RESERVED_TOKENS = {
 439        *Tokenizer.SINGLE_TOKENS.values(),
 440        TokenType.SELECT,
 441    } - {TokenType.IDENTIFIER}
 442
 443    DB_CREATABLES = {
 444        TokenType.DATABASE,
 445        TokenType.DICTIONARY,
 446        TokenType.FILE_FORMAT,
 447        TokenType.MODEL,
 448        TokenType.NAMESPACE,
 449        TokenType.SCHEMA,
 450        TokenType.SEQUENCE,
 451        TokenType.SINK,
 452        TokenType.SOURCE,
 453        TokenType.STAGE,
 454        TokenType.STORAGE_INTEGRATION,
 455        TokenType.STREAMLIT,
 456        TokenType.TABLE,
 457        TokenType.TAG,
 458        TokenType.VIEW,
 459        TokenType.WAREHOUSE,
 460    }
 461
 462    CREATABLES = {
 463        TokenType.COLUMN,
 464        TokenType.CONSTRAINT,
 465        TokenType.FOREIGN_KEY,
 466        TokenType.FUNCTION,
 467        TokenType.INDEX,
 468        TokenType.PROCEDURE,
 469        *DB_CREATABLES,
 470    }
 471
 472    ALTERABLES = {
 473        TokenType.INDEX,
 474        TokenType.TABLE,
 475        TokenType.VIEW,
 476    }
 477
 478    # Tokens that can represent identifiers
 479    ID_VAR_TOKENS = {
 480        TokenType.ALL,
 481        TokenType.ATTACH,
 482        TokenType.VAR,
 483        TokenType.ANTI,
 484        TokenType.APPLY,
 485        TokenType.ASC,
 486        TokenType.ASOF,
 487        TokenType.AUTO_INCREMENT,
 488        TokenType.BEGIN,
 489        TokenType.BPCHAR,
 490        TokenType.CACHE,
 491        TokenType.CASE,
 492        TokenType.COLLATE,
 493        TokenType.COMMAND,
 494        TokenType.COMMENT,
 495        TokenType.COMMIT,
 496        TokenType.CONSTRAINT,
 497        TokenType.COPY,
 498        TokenType.CUBE,
 499        TokenType.CURRENT_SCHEMA,
 500        TokenType.DEFAULT,
 501        TokenType.DELETE,
 502        TokenType.DESC,
 503        TokenType.DESCRIBE,
 504        TokenType.DETACH,
 505        TokenType.DICTIONARY,
 506        TokenType.DIV,
 507        TokenType.END,
 508        TokenType.EXECUTE,
 509        TokenType.EXPORT,
 510        TokenType.ESCAPE,
 511        TokenType.FALSE,
 512        TokenType.FIRST,
 513        TokenType.FILTER,
 514        TokenType.FINAL,
 515        TokenType.FORMAT,
 516        TokenType.FULL,
 517        TokenType.GET,
 518        TokenType.IDENTIFIER,
 519        TokenType.IS,
 520        TokenType.ISNULL,
 521        TokenType.INTERVAL,
 522        TokenType.KEEP,
 523        TokenType.KILL,
 524        TokenType.LEFT,
 525        TokenType.LIMIT,
 526        TokenType.LOAD,
 527        TokenType.MERGE,
 528        TokenType.NATURAL,
 529        TokenType.NEXT,
 530        TokenType.OFFSET,
 531        TokenType.OPERATOR,
 532        TokenType.ORDINALITY,
 533        TokenType.OVERLAPS,
 534        TokenType.OVERWRITE,
 535        TokenType.PARTITION,
 536        TokenType.PERCENT,
 537        TokenType.PIVOT,
 538        TokenType.PRAGMA,
 539        TokenType.PUT,
 540        TokenType.RANGE,
 541        TokenType.RECURSIVE,
 542        TokenType.REFERENCES,
 543        TokenType.REFRESH,
 544        TokenType.RENAME,
 545        TokenType.REPLACE,
 546        TokenType.RIGHT,
 547        TokenType.ROLLUP,
 548        TokenType.ROW,
 549        TokenType.ROWS,
 550        TokenType.SEMI,
 551        TokenType.SET,
 552        TokenType.SETTINGS,
 553        TokenType.SHOW,
 554        TokenType.TEMPORARY,
 555        TokenType.TOP,
 556        TokenType.TRUE,
 557        TokenType.TRUNCATE,
 558        TokenType.UNIQUE,
 559        TokenType.UNNEST,
 560        TokenType.UNPIVOT,
 561        TokenType.UPDATE,
 562        TokenType.USE,
 563        TokenType.VOLATILE,
 564        TokenType.WINDOW,
 565        *CREATABLES,
 566        *SUBQUERY_PREDICATES,
 567        *TYPE_TOKENS,
 568        *NO_PAREN_FUNCTIONS,
 569    }
 570    ID_VAR_TOKENS.remove(TokenType.UNION)
 571
 572    TABLE_ALIAS_TOKENS = ID_VAR_TOKENS - {
 573        TokenType.ANTI,
 574        TokenType.APPLY,
 575        TokenType.ASOF,
 576        TokenType.FULL,
 577        TokenType.LEFT,
 578        TokenType.LOCK,
 579        TokenType.NATURAL,
 580        TokenType.RIGHT,
 581        TokenType.SEMI,
 582        TokenType.WINDOW,
 583    }
 584
 585    ALIAS_TOKENS = ID_VAR_TOKENS
 586
 587    COLON_PLACEHOLDER_TOKENS = ID_VAR_TOKENS
 588
 589    ARRAY_CONSTRUCTORS = {
 590        "ARRAY": exp.Array,
 591        "LIST": exp.List,
 592    }
 593
 594    COMMENT_TABLE_ALIAS_TOKENS = TABLE_ALIAS_TOKENS - {TokenType.IS}
 595
 596    UPDATE_ALIAS_TOKENS = TABLE_ALIAS_TOKENS - {TokenType.SET}
 597
 598    TRIM_TYPES = {"LEADING", "TRAILING", "BOTH"}
 599
 600    FUNC_TOKENS = {
 601        TokenType.COLLATE,
 602        TokenType.COMMAND,
 603        TokenType.CURRENT_DATE,
 604        TokenType.CURRENT_DATETIME,
 605        TokenType.CURRENT_SCHEMA,
 606        TokenType.CURRENT_TIMESTAMP,
 607        TokenType.CURRENT_TIME,
 608        TokenType.CURRENT_USER,
 609        TokenType.FILTER,
 610        TokenType.FIRST,
 611        TokenType.FORMAT,
 612        TokenType.GET,
 613        TokenType.GLOB,
 614        TokenType.IDENTIFIER,
 615        TokenType.INDEX,
 616        TokenType.ISNULL,
 617        TokenType.ILIKE,
 618        TokenType.INSERT,
 619        TokenType.LIKE,
 620        TokenType.MERGE,
 621        TokenType.NEXT,
 622        TokenType.OFFSET,
 623        TokenType.PRIMARY_KEY,
 624        TokenType.RANGE,
 625        TokenType.REPLACE,
 626        TokenType.RLIKE,
 627        TokenType.ROW,
 628        TokenType.UNNEST,
 629        TokenType.VAR,
 630        TokenType.LEFT,
 631        TokenType.RIGHT,
 632        TokenType.SEQUENCE,
 633        TokenType.DATE,
 634        TokenType.DATETIME,
 635        TokenType.TABLE,
 636        TokenType.TIMESTAMP,
 637        TokenType.TIMESTAMPTZ,
 638        TokenType.TRUNCATE,
 639        TokenType.WINDOW,
 640        TokenType.XOR,
 641        *TYPE_TOKENS,
 642        *SUBQUERY_PREDICATES,
 643    }
 644
 645    CONJUNCTION: t.Dict[TokenType, t.Type[exp.Expression]] = {
 646        TokenType.AND: exp.And,
 647    }
 648
 649    ASSIGNMENT: t.Dict[TokenType, t.Type[exp.Expression]] = {
 650        TokenType.COLON_EQ: exp.PropertyEQ,
 651    }
 652
 653    DISJUNCTION: t.Dict[TokenType, t.Type[exp.Expression]] = {
 654        TokenType.OR: exp.Or,
 655    }
 656
 657    EQUALITY = {
 658        TokenType.EQ: exp.EQ,
 659        TokenType.NEQ: exp.NEQ,
 660        TokenType.NULLSAFE_EQ: exp.NullSafeEQ,
 661    }
 662
 663    COMPARISON = {
 664        TokenType.GT: exp.GT,
 665        TokenType.GTE: exp.GTE,
 666        TokenType.LT: exp.LT,
 667        TokenType.LTE: exp.LTE,
 668    }
 669
 670    BITWISE = {
 671        TokenType.AMP: exp.BitwiseAnd,
 672        TokenType.CARET: exp.BitwiseXor,
 673        TokenType.PIPE: exp.BitwiseOr,
 674    }
 675
 676    TERM = {
 677        TokenType.DASH: exp.Sub,
 678        TokenType.PLUS: exp.Add,
 679        TokenType.MOD: exp.Mod,
 680        TokenType.COLLATE: exp.Collate,
 681    }
 682
 683    FACTOR = {
 684        TokenType.DIV: exp.IntDiv,
 685        TokenType.LR_ARROW: exp.Distance,
 686        TokenType.SLASH: exp.Div,
 687        TokenType.STAR: exp.Mul,
 688    }
 689
 690    EXPONENT: t.Dict[TokenType, t.Type[exp.Expression]] = {}
 691
 692    TIMES = {
 693        TokenType.TIME,
 694        TokenType.TIMETZ,
 695    }
 696
 697    TIMESTAMPS = {
 698        TokenType.TIMESTAMP,
 699        TokenType.TIMESTAMPNTZ,
 700        TokenType.TIMESTAMPTZ,
 701        TokenType.TIMESTAMPLTZ,
 702        *TIMES,
 703    }
 704
 705    SET_OPERATIONS = {
 706        TokenType.UNION,
 707        TokenType.INTERSECT,
 708        TokenType.EXCEPT,
 709    }
 710
 711    JOIN_METHODS = {
 712        TokenType.ASOF,
 713        TokenType.NATURAL,
 714        TokenType.POSITIONAL,
 715    }
 716
 717    JOIN_SIDES = {
 718        TokenType.LEFT,
 719        TokenType.RIGHT,
 720        TokenType.FULL,
 721    }
 722
 723    JOIN_KINDS = {
 724        TokenType.ANTI,
 725        TokenType.CROSS,
 726        TokenType.INNER,
 727        TokenType.OUTER,
 728        TokenType.SEMI,
 729        TokenType.STRAIGHT_JOIN,
 730    }
 731
 732    JOIN_HINTS: t.Set[str] = set()
 733
 734    LAMBDAS = {
 735        TokenType.ARROW: lambda self, expressions: self.expression(
 736            exp.Lambda,
 737            this=self._replace_lambda(
 738                self._parse_assignment(),
 739                expressions,
 740            ),
 741            expressions=expressions,
 742        ),
 743        TokenType.FARROW: lambda self, expressions: self.expression(
 744            exp.Kwarg,
 745            this=exp.var(expressions[0].name),
 746            expression=self._parse_assignment(),
 747        ),
 748    }
 749
 750    COLUMN_OPERATORS = {
 751        TokenType.DOT: None,
 752        TokenType.DOTCOLON: lambda self, this, to: self.expression(
 753            exp.JSONCast,
 754            this=this,
 755            to=to,
 756        ),
 757        TokenType.DCOLON: lambda self, this, to: self.expression(
 758            exp.Cast if self.STRICT_CAST else exp.TryCast,
 759            this=this,
 760            to=to,
 761        ),
 762        TokenType.ARROW: lambda self, this, path: self.expression(
 763            exp.JSONExtract,
 764            this=this,
 765            expression=self.dialect.to_json_path(path),
 766            only_json_types=self.JSON_ARROWS_REQUIRE_JSON_TYPE,
 767        ),
 768        TokenType.DARROW: lambda self, this, path: self.expression(
 769            exp.JSONExtractScalar,
 770            this=this,
 771            expression=self.dialect.to_json_path(path),
 772            only_json_types=self.JSON_ARROWS_REQUIRE_JSON_TYPE,
 773        ),
 774        TokenType.HASH_ARROW: lambda self, this, path: self.expression(
 775            exp.JSONBExtract,
 776            this=this,
 777            expression=path,
 778        ),
 779        TokenType.DHASH_ARROW: lambda self, this, path: self.expression(
 780            exp.JSONBExtractScalar,
 781            this=this,
 782            expression=path,
 783        ),
 784        TokenType.PLACEHOLDER: lambda self, this, key: self.expression(
 785            exp.JSONBContains,
 786            this=this,
 787            expression=key,
 788        ),
 789    }
 790
 791    EXPRESSION_PARSERS = {
 792        exp.Cluster: lambda self: self._parse_sort(exp.Cluster, TokenType.CLUSTER_BY),
 793        exp.Column: lambda self: self._parse_column(),
 794        exp.Condition: lambda self: self._parse_assignment(),
 795        exp.DataType: lambda self: self._parse_types(allow_identifiers=False, schema=True),
 796        exp.Expression: lambda self: self._parse_expression(),
 797        exp.From: lambda self: self._parse_from(joins=True),
 798        exp.Group: lambda self: self._parse_group(),
 799        exp.Having: lambda self: self._parse_having(),
 800        exp.Hint: lambda self: self._parse_hint_body(),
 801        exp.Identifier: lambda self: self._parse_id_var(),
 802        exp.Join: lambda self: self._parse_join(),
 803        exp.Lambda: lambda self: self._parse_lambda(),
 804        exp.Lateral: lambda self: self._parse_lateral(),
 805        exp.Limit: lambda self: self._parse_limit(),
 806        exp.Offset: lambda self: self._parse_offset(),
 807        exp.Order: lambda self: self._parse_order(),
 808        exp.Ordered: lambda self: self._parse_ordered(),
 809        exp.Properties: lambda self: self._parse_properties(),
 810        exp.PartitionedByProperty: lambda self: self._parse_partitioned_by(),
 811        exp.Qualify: lambda self: self._parse_qualify(),
 812        exp.Returning: lambda self: self._parse_returning(),
 813        exp.Select: lambda self: self._parse_select(),
 814        exp.Sort: lambda self: self._parse_sort(exp.Sort, TokenType.SORT_BY),
 815        exp.Table: lambda self: self._parse_table_parts(),
 816        exp.TableAlias: lambda self: self._parse_table_alias(),
 817        exp.Tuple: lambda self: self._parse_value(values=False),
 818        exp.Whens: lambda self: self._parse_when_matched(),
 819        exp.Where: lambda self: self._parse_where(),
 820        exp.Window: lambda self: self._parse_named_window(),
 821        exp.With: lambda self: self._parse_with(),
 822        "JOIN_TYPE": lambda self: self._parse_join_parts(),
 823    }
 824
 825    STATEMENT_PARSERS = {
 826        TokenType.ALTER: lambda self: self._parse_alter(),
 827        TokenType.ANALYZE: lambda self: self._parse_analyze(),
 828        TokenType.BEGIN: lambda self: self._parse_transaction(),
 829        TokenType.CACHE: lambda self: self._parse_cache(),
 830        TokenType.COMMENT: lambda self: self._parse_comment(),
 831        TokenType.COMMIT: lambda self: self._parse_commit_or_rollback(),
 832        TokenType.COPY: lambda self: self._parse_copy(),
 833        TokenType.CREATE: lambda self: self._parse_create(),
 834        TokenType.DELETE: lambda self: self._parse_delete(),
 835        TokenType.DESC: lambda self: self._parse_describe(),
 836        TokenType.DESCRIBE: lambda self: self._parse_describe(),
 837        TokenType.DROP: lambda self: self._parse_drop(),
 838        TokenType.GRANT: lambda self: self._parse_grant(),
 839        TokenType.INSERT: lambda self: self._parse_insert(),
 840        TokenType.KILL: lambda self: self._parse_kill(),
 841        TokenType.LOAD: lambda self: self._parse_load(),
 842        TokenType.MERGE: lambda self: self._parse_merge(),
 843        TokenType.PIVOT: lambda self: self._parse_simplified_pivot(),
 844        TokenType.PRAGMA: lambda self: self.expression(exp.Pragma, this=self._parse_expression()),
 845        TokenType.REFRESH: lambda self: self._parse_refresh(),
 846        TokenType.ROLLBACK: lambda self: self._parse_commit_or_rollback(),
 847        TokenType.SET: lambda self: self._parse_set(),
 848        TokenType.TRUNCATE: lambda self: self._parse_truncate_table(),
 849        TokenType.UNCACHE: lambda self: self._parse_uncache(),
 850        TokenType.UNPIVOT: lambda self: self._parse_simplified_pivot(is_unpivot=True),
 851        TokenType.UPDATE: lambda self: self._parse_update(),
 852        TokenType.USE: lambda self: self._parse_use(),
 853        TokenType.SEMICOLON: lambda self: exp.Semicolon(),
 854    }
 855
 856    UNARY_PARSERS = {
 857        TokenType.PLUS: lambda self: self._parse_unary(),  # Unary + is handled as a no-op
 858        TokenType.NOT: lambda self: self.expression(exp.Not, this=self._parse_equality()),
 859        TokenType.TILDA: lambda self: self.expression(exp.BitwiseNot, this=self._parse_unary()),
 860        TokenType.DASH: lambda self: self.expression(exp.Neg, this=self._parse_unary()),
 861        TokenType.PIPE_SLASH: lambda self: self.expression(exp.Sqrt, this=self._parse_unary()),
 862        TokenType.DPIPE_SLASH: lambda self: self.expression(exp.Cbrt, this=self._parse_unary()),
 863    }
 864
 865    STRING_PARSERS = {
 866        TokenType.HEREDOC_STRING: lambda self, token: self.expression(
 867            exp.RawString, this=token.text
 868        ),
 869        TokenType.NATIONAL_STRING: lambda self, token: self.expression(
 870            exp.National, this=token.text
 871        ),
 872        TokenType.RAW_STRING: lambda self, token: self.expression(exp.RawString, this=token.text),
 873        TokenType.STRING: lambda self, token: self.expression(
 874            exp.Literal, this=token.text, is_string=True
 875        ),
 876        TokenType.UNICODE_STRING: lambda self, token: self.expression(
 877            exp.UnicodeString,
 878            this=token.text,
 879            escape=self._match_text_seq("UESCAPE") and self._parse_string(),
 880        ),
 881    }
 882
 883    NUMERIC_PARSERS = {
 884        TokenType.BIT_STRING: lambda self, token: self.expression(exp.BitString, this=token.text),
 885        TokenType.BYTE_STRING: lambda self, token: self.expression(exp.ByteString, this=token.text),
 886        TokenType.HEX_STRING: lambda self, token: self.expression(
 887            exp.HexString,
 888            this=token.text,
 889            is_integer=self.dialect.HEX_STRING_IS_INTEGER_TYPE or None,
 890        ),
 891        TokenType.NUMBER: lambda self, token: self.expression(
 892            exp.Literal, this=token.text, is_string=False
 893        ),
 894    }
 895
 896    PRIMARY_PARSERS = {
 897        **STRING_PARSERS,
 898        **NUMERIC_PARSERS,
 899        TokenType.INTRODUCER: lambda self, token: self._parse_introducer(token),
 900        TokenType.NULL: lambda self, _: self.expression(exp.Null),
 901        TokenType.TRUE: lambda self, _: self.expression(exp.Boolean, this=True),
 902        TokenType.FALSE: lambda self, _: self.expression(exp.Boolean, this=False),
 903        TokenType.SESSION_PARAMETER: lambda self, _: self._parse_session_parameter(),
 904        TokenType.STAR: lambda self, _: self._parse_star_ops(),
 905    }
 906
 907    PLACEHOLDER_PARSERS = {
 908        TokenType.PLACEHOLDER: lambda self: self.expression(exp.Placeholder),
 909        TokenType.PARAMETER: lambda self: self._parse_parameter(),
 910        TokenType.COLON: lambda self: (
 911            self.expression(exp.Placeholder, this=self._prev.text)
 912            if self._match_set(self.COLON_PLACEHOLDER_TOKENS)
 913            else None
 914        ),
 915    }
 916
 917    RANGE_PARSERS = {
 918        TokenType.AT_GT: binary_range_parser(exp.ArrayContainsAll),
 919        TokenType.BETWEEN: lambda self, this: self._parse_between(this),
 920        TokenType.GLOB: binary_range_parser(exp.Glob),
 921        TokenType.ILIKE: binary_range_parser(exp.ILike),
 922        TokenType.IN: lambda self, this: self._parse_in(this),
 923        TokenType.IRLIKE: binary_range_parser(exp.RegexpILike),
 924        TokenType.IS: lambda self, this: self._parse_is(this),
 925        TokenType.LIKE: binary_range_parser(exp.Like),
 926        TokenType.LT_AT: binary_range_parser(exp.ArrayContainsAll, reverse_args=True),
 927        TokenType.OVERLAPS: binary_range_parser(exp.Overlaps),
 928        TokenType.RLIKE: binary_range_parser(exp.RegexpLike),
 929        TokenType.SIMILAR_TO: binary_range_parser(exp.SimilarTo),
 930        TokenType.FOR: lambda self, this: self._parse_comprehension(this),
 931    }
 932
 933    PROPERTY_PARSERS: t.Dict[str, t.Callable] = {
 934        "ALLOWED_VALUES": lambda self: self.expression(
 935            exp.AllowedValuesProperty, expressions=self._parse_csv(self._parse_primary)
 936        ),
 937        "ALGORITHM": lambda self: self._parse_property_assignment(exp.AlgorithmProperty),
 938        "AUTO": lambda self: self._parse_auto_property(),
 939        "AUTO_INCREMENT": lambda self: self._parse_property_assignment(exp.AutoIncrementProperty),
 940        "BACKUP": lambda self: self.expression(
 941            exp.BackupProperty, this=self._parse_var(any_token=True)
 942        ),
 943        "BLOCKCOMPRESSION": lambda self: self._parse_blockcompression(),
 944        "CHARSET": lambda self, **kwargs: self._parse_character_set(**kwargs),
 945        "CHARACTER SET": lambda self, **kwargs: self._parse_character_set(**kwargs),
 946        "CHECKSUM": lambda self: self._parse_checksum(),
 947        "CLUSTER BY": lambda self: self._parse_cluster(),
 948        "CLUSTERED": lambda self: self._parse_clustered_by(),
 949        "COLLATE": lambda self, **kwargs: self._parse_property_assignment(
 950            exp.CollateProperty, **kwargs
 951        ),
 952        "COMMENT": lambda self: self._parse_property_assignment(exp.SchemaCommentProperty),
 953        "CONTAINS": lambda self: self._parse_contains_property(),
 954        "COPY": lambda self: self._parse_copy_property(),
 955        "DATABLOCKSIZE": lambda self, **kwargs: self._parse_datablocksize(**kwargs),
 956        "DATA_DELETION": lambda self: self._parse_data_deletion_property(),
 957        "DEFINER": lambda self: self._parse_definer(),
 958        "DETERMINISTIC": lambda self: self.expression(
 959            exp.StabilityProperty, this=exp.Literal.string("IMMUTABLE")
 960        ),
 961        "DISTRIBUTED": lambda self: self._parse_distributed_property(),
 962        "DUPLICATE": lambda self: self._parse_composite_key_property(exp.DuplicateKeyProperty),
 963        "DYNAMIC": lambda self: self.expression(exp.DynamicProperty),
 964        "DISTKEY": lambda self: self._parse_distkey(),
 965        "DISTSTYLE": lambda self: self._parse_property_assignment(exp.DistStyleProperty),
 966        "EMPTY": lambda self: self.expression(exp.EmptyProperty),
 967        "ENGINE": lambda self: self._parse_property_assignment(exp.EngineProperty),
 968        "ENVIRONMENT": lambda self: self.expression(
 969            exp.EnviromentProperty, expressions=self._parse_wrapped_csv(self._parse_assignment)
 970        ),
 971        "EXECUTE": lambda self: self._parse_property_assignment(exp.ExecuteAsProperty),
 972        "EXTERNAL": lambda self: self.expression(exp.ExternalProperty),
 973        "FALLBACK": lambda self, **kwargs: self._parse_fallback(**kwargs),
 974        "FORMAT": lambda self: self._parse_property_assignment(exp.FileFormatProperty),
 975        "FREESPACE": lambda self: self._parse_freespace(),
 976        "GLOBAL": lambda self: self.expression(exp.GlobalProperty),
 977        "HEAP": lambda self: self.expression(exp.HeapProperty),
 978        "ICEBERG": lambda self: self.expression(exp.IcebergProperty),
 979        "IMMUTABLE": lambda self: self.expression(
 980            exp.StabilityProperty, this=exp.Literal.string("IMMUTABLE")
 981        ),
 982        "INHERITS": lambda self: self.expression(
 983            exp.InheritsProperty, expressions=self._parse_wrapped_csv(self._parse_table)
 984        ),
 985        "INPUT": lambda self: self.expression(exp.InputModelProperty, this=self._parse_schema()),
 986        "JOURNAL": lambda self, **kwargs: self._parse_journal(**kwargs),
 987        "LANGUAGE": lambda self: self._parse_property_assignment(exp.LanguageProperty),
 988        "LAYOUT": lambda self: self._parse_dict_property(this="LAYOUT"),
 989        "LIFETIME": lambda self: self._parse_dict_range(this="LIFETIME"),
 990        "LIKE": lambda self: self._parse_create_like(),
 991        "LOCATION": lambda self: self._parse_property_assignment(exp.LocationProperty),
 992        "LOCK": lambda self: self._parse_locking(),
 993        "LOCKING": lambda self: self._parse_locking(),
 994        "LOG": lambda self, **kwargs: self._parse_log(**kwargs),
 995        "MATERIALIZED": lambda self: self.expression(exp.MaterializedProperty),
 996        "MERGEBLOCKRATIO": lambda self, **kwargs: self._parse_mergeblockratio(**kwargs),
 997        "MODIFIES": lambda self: self._parse_modifies_property(),
 998        "MULTISET": lambda self: self.expression(exp.SetProperty, multi=True),
 999        "NO": lambda self: self._parse_no_property(),
1000        "ON": lambda self: self._parse_on_property(),
1001        "ORDER BY": lambda self: self._parse_order(skip_order_token=True),
1002        "OUTPUT": lambda self: self.expression(exp.OutputModelProperty, this=self._parse_schema()),
1003        "PARTITION": lambda self: self._parse_partitioned_of(),
1004        "PARTITION BY": lambda self: self._parse_partitioned_by(),
1005        "PARTITIONED BY": lambda self: self._parse_partitioned_by(),
1006        "PARTITIONED_BY": lambda self: self._parse_partitioned_by(),
1007        "PRIMARY KEY": lambda self: self._parse_primary_key(in_props=True),
1008        "RANGE": lambda self: self._parse_dict_range(this="RANGE"),
1009        "READS": lambda self: self._parse_reads_property(),
1010        "REMOTE": lambda self: self._parse_remote_with_connection(),
1011        "RETURNS": lambda self: self._parse_returns(),
1012        "STRICT": lambda self: self.expression(exp.StrictProperty),
1013        "STREAMING": lambda self: self.expression(exp.StreamingTableProperty),
1014        "ROW": lambda self: self._parse_row(),
1015        "ROW_FORMAT": lambda self: self._parse_property_assignment(exp.RowFormatProperty),
1016        "SAMPLE": lambda self: self.expression(
1017            exp.SampleProperty, this=self._match_text_seq("BY") and self._parse_bitwise()
1018        ),
1019        "SECURE": lambda self: self.expression(exp.SecureProperty),
1020        "SECURITY": lambda self: self._parse_security(),
1021        "SET": lambda self: self.expression(exp.SetProperty, multi=False),
1022        "SETTINGS": lambda self: self._parse_settings_property(),
1023        "SHARING": lambda self: self._parse_property_assignment(exp.SharingProperty),
1024        "SORTKEY": lambda self: self._parse_sortkey(),
1025        "SOURCE": lambda self: self._parse_dict_property(this="SOURCE"),
1026        "STABLE": lambda self: self.expression(
1027            exp.StabilityProperty, this=exp.Literal.string("STABLE")
1028        ),
1029        "STORED": lambda self: self._parse_stored(),
1030        "SYSTEM_VERSIONING": lambda self: self._parse_system_versioning_property(),
1031        "TBLPROPERTIES": lambda self: self._parse_wrapped_properties(),
1032        "TEMP": lambda self: self.expression(exp.TemporaryProperty),
1033        "TEMPORARY": lambda self: self.expression(exp.TemporaryProperty),
1034        "TO": lambda self: self._parse_to_table(),
1035        "TRANSIENT": lambda self: self.expression(exp.TransientProperty),
1036        "TRANSFORM": lambda self: self.expression(
1037            exp.TransformModelProperty, expressions=self._parse_wrapped_csv(self._parse_expression)
1038        ),
1039        "TTL": lambda self: self._parse_ttl(),
1040        "USING": lambda self: self._parse_property_assignment(exp.FileFormatProperty),
1041        "UNLOGGED": lambda self: self.expression(exp.UnloggedProperty),
1042        "VOLATILE": lambda self: self._parse_volatile_property(),
1043        "WITH": lambda self: self._parse_with_property(),
1044    }
1045
1046    CONSTRAINT_PARSERS = {
1047        "AUTOINCREMENT": lambda self: self._parse_auto_increment(),
1048        "AUTO_INCREMENT": lambda self: self._parse_auto_increment(),
1049        "CASESPECIFIC": lambda self: self.expression(exp.CaseSpecificColumnConstraint, not_=False),
1050        "CHARACTER SET": lambda self: self.expression(
1051            exp.CharacterSetColumnConstraint, this=self._parse_var_or_string()
1052        ),
1053        "CHECK": lambda self: self.expression(
1054            exp.CheckColumnConstraint,
1055            this=self._parse_wrapped(self._parse_assignment),
1056            enforced=self._match_text_seq("ENFORCED"),
1057        ),
1058        "COLLATE": lambda self: self.expression(
1059            exp.CollateColumnConstraint,
1060            this=self._parse_identifier() or self._parse_column(),
1061        ),
1062        "COMMENT": lambda self: self.expression(
1063            exp.CommentColumnConstraint, this=self._parse_string()
1064        ),
1065        "COMPRESS": lambda self: self._parse_compress(),
1066        "CLUSTERED": lambda self: self.expression(
1067            exp.ClusteredColumnConstraint, this=self._parse_wrapped_csv(self._parse_ordered)
1068        ),
1069        "NONCLUSTERED": lambda self: self.expression(
1070            exp.NonClusteredColumnConstraint, this=self._parse_wrapped_csv(self._parse_ordered)
1071        ),
1072        "DEFAULT": lambda self: self.expression(
1073            exp.DefaultColumnConstraint, this=self._parse_bitwise()
1074        ),
1075        "ENCODE": lambda self: self.expression(exp.EncodeColumnConstraint, this=self._parse_var()),
1076        "EPHEMERAL": lambda self: self.expression(
1077            exp.EphemeralColumnConstraint, this=self._parse_bitwise()
1078        ),
1079        "EXCLUDE": lambda self: self.expression(
1080            exp.ExcludeColumnConstraint, this=self._parse_index_params()
1081        ),
1082        "FOREIGN KEY": lambda self: self._parse_foreign_key(),
1083        "FORMAT": lambda self: self.expression(
1084            exp.DateFormatColumnConstraint, this=self._parse_var_or_string()
1085        ),
1086        "GENERATED": lambda self: self._parse_generated_as_identity(),
1087        "IDENTITY": lambda self: self._parse_auto_increment(),
1088        "INLINE": lambda self: self._parse_inline(),
1089        "LIKE": lambda self: self._parse_create_like(),
1090        "NOT": lambda self: self._parse_not_constraint(),
1091        "NULL": lambda self: self.expression(exp.NotNullColumnConstraint, allow_null=True),
1092        "ON": lambda self: (
1093            self._match(TokenType.UPDATE)
1094            and self.expression(exp.OnUpdateColumnConstraint, this=self._parse_function())
1095        )
1096        or self.expression(exp.OnProperty, this=self._parse_id_var()),
1097        "PATH": lambda self: self.expression(exp.PathColumnConstraint, this=self._parse_string()),
1098        "PERIOD": lambda self: self._parse_period_for_system_time(),
1099        "PRIMARY KEY": lambda self: self._parse_primary_key(),
1100        "REFERENCES": lambda self: self._parse_references(match=False),
1101        "TITLE": lambda self: self.expression(
1102            exp.TitleColumnConstraint, this=self._parse_var_or_string()
1103        ),
1104        "TTL": lambda self: self.expression(exp.MergeTreeTTL, expressions=[self._parse_bitwise()]),
1105        "UNIQUE": lambda self: self._parse_unique(),
1106        "UPPERCASE": lambda self: self.expression(exp.UppercaseColumnConstraint),
1107        "WATERMARK": lambda self: self.expression(
1108            exp.WatermarkColumnConstraint,
1109            this=self._match(TokenType.FOR) and self._parse_column(),
1110            expression=self._match(TokenType.ALIAS) and self._parse_disjunction(),
1111        ),
1112        "WITH": lambda self: self.expression(
1113            exp.Properties, expressions=self._parse_wrapped_properties()
1114        ),
1115        "BUCKET": lambda self: self._parse_partitioned_by_bucket_or_truncate(),
1116        "TRUNCATE": lambda self: self._parse_partitioned_by_bucket_or_truncate(),
1117    }
1118
1119    def _parse_partitioned_by_bucket_or_truncate(self) -> exp.Expression:
1120        klass = (
1121            exp.PartitionedByBucket
1122            if self._prev.text.upper() == "BUCKET"
1123            else exp.PartitionByTruncate
1124        )
1125
1126        args = self._parse_wrapped_csv(lambda: self._parse_primary() or self._parse_column())
1127        this, expression = seq_get(args, 0), seq_get(args, 1)
1128
1129        if isinstance(this, exp.Literal):
1130            # Check for Iceberg partition transforms (bucket / truncate) and ensure their arguments are in the right order
1131            #  - For Hive, it's `bucket(<num buckets>, <col name>)` or `truncate(<num_chars>, <col_name>)`
1132            #  - For Trino, it's reversed - `bucket(<col name>, <num buckets>)` or `truncate(<col_name>, <num_chars>)`
1133            # Both variants are canonicalized in the latter i.e `bucket(<col name>, <num buckets>)`
1134            #
1135            # Hive ref: https://docs.aws.amazon.com/athena/latest/ug/querying-iceberg-creating-tables.html#querying-iceberg-partitioning
1136            # Trino ref: https://docs.aws.amazon.com/athena/latest/ug/create-table-as.html#ctas-table-properties
1137            this, expression = expression, this
1138
1139        return self.expression(klass, this=this, expression=expression)
1140
1141    ALTER_PARSERS = {
1142        "ADD": lambda self: self._parse_alter_table_add(),
1143        "AS": lambda self: self._parse_select(),
1144        "ALTER": lambda self: self._parse_alter_table_alter(),
1145        "CLUSTER BY": lambda self: self._parse_cluster(wrapped=True),
1146        "DELETE": lambda self: self.expression(exp.Delete, where=self._parse_where()),
1147        "DROP": lambda self: self._parse_alter_table_drop(),
1148        "RENAME": lambda self: self._parse_alter_table_rename(),
1149        "SET": lambda self: self._parse_alter_table_set(),
1150        "SWAP": lambda self: self.expression(
1151            exp.SwapTable, this=self._match(TokenType.WITH) and self._parse_table(schema=True)
1152        ),
1153    }
1154
1155    ALTER_ALTER_PARSERS = {
1156        "DISTKEY": lambda self: self._parse_alter_diststyle(),
1157        "DISTSTYLE": lambda self: self._parse_alter_diststyle(),
1158        "SORTKEY": lambda self: self._parse_alter_sortkey(),
1159        "COMPOUND": lambda self: self._parse_alter_sortkey(compound=True),
1160    }
1161
1162    SCHEMA_UNNAMED_CONSTRAINTS = {
1163        "CHECK",
1164        "EXCLUDE",
1165        "FOREIGN KEY",
1166        "LIKE",
1167        "PERIOD",
1168        "PRIMARY KEY",
1169        "UNIQUE",
1170        "WATERMARK",
1171        "BUCKET",
1172        "TRUNCATE",
1173    }
1174
1175    NO_PAREN_FUNCTION_PARSERS = {
1176        "ANY": lambda self: self.expression(exp.Any, this=self._parse_bitwise()),
1177        "CASE": lambda self: self._parse_case(),
1178        "CONNECT_BY_ROOT": lambda self: self.expression(
1179            exp.ConnectByRoot, this=self._parse_column()
1180        ),
1181        "IF": lambda self: self._parse_if(),
1182    }
1183
1184    INVALID_FUNC_NAME_TOKENS = {
1185        TokenType.IDENTIFIER,
1186        TokenType.STRING,
1187    }
1188
1189    FUNCTIONS_WITH_ALIASED_ARGS = {"STRUCT"}
1190
1191    KEY_VALUE_DEFINITIONS = (exp.Alias, exp.EQ, exp.PropertyEQ, exp.Slice)
1192
1193    FUNCTION_PARSERS = {
1194        **{
1195            name: lambda self: self._parse_max_min_by(exp.ArgMax) for name in exp.ArgMax.sql_names()
1196        },
1197        **{
1198            name: lambda self: self._parse_max_min_by(exp.ArgMin) for name in exp.ArgMin.sql_names()
1199        },
1200        "CAST": lambda self: self._parse_cast(self.STRICT_CAST),
1201        "CEIL": lambda self: self._parse_ceil_floor(exp.Ceil),
1202        "CONVERT": lambda self: self._parse_convert(self.STRICT_CAST),
1203        "DECODE": lambda self: self._parse_decode(),
1204        "EXTRACT": lambda self: self._parse_extract(),
1205        "FLOOR": lambda self: self._parse_ceil_floor(exp.Floor),
1206        "GAP_FILL": lambda self: self._parse_gap_fill(),
1207        "JSON_OBJECT": lambda self: self._parse_json_object(),
1208        "JSON_OBJECTAGG": lambda self: self._parse_json_object(agg=True),
1209        "JSON_TABLE": lambda self: self._parse_json_table(),
1210        "MATCH": lambda self: self._parse_match_against(),
1211        "NORMALIZE": lambda self: self._parse_normalize(),
1212        "OPENJSON": lambda self: self._parse_open_json(),
1213        "OVERLAY": lambda self: self._parse_overlay(),
1214        "POSITION": lambda self: self._parse_position(),
1215        "PREDICT": lambda self: self._parse_predict(),
1216        "SAFE_CAST": lambda self: self._parse_cast(False, safe=True),
1217        "STRING_AGG": lambda self: self._parse_string_agg(),
1218        "SUBSTRING": lambda self: self._parse_substring(),
1219        "TRIM": lambda self: self._parse_trim(),
1220        "TRY_CAST": lambda self: self._parse_cast(False, safe=True),
1221        "TRY_CONVERT": lambda self: self._parse_convert(False, safe=True),
1222        "XMLELEMENT": lambda self: self.expression(
1223            exp.XMLElement,
1224            this=self._match_text_seq("NAME") and self._parse_id_var(),
1225            expressions=self._match(TokenType.COMMA) and self._parse_csv(self._parse_expression),
1226        ),
1227        "XMLTABLE": lambda self: self._parse_xml_table(),
1228    }
1229
1230    QUERY_MODIFIER_PARSERS = {
1231        TokenType.MATCH_RECOGNIZE: lambda self: ("match", self._parse_match_recognize()),
1232        TokenType.PREWHERE: lambda self: ("prewhere", self._parse_prewhere()),
1233        TokenType.WHERE: lambda self: ("where", self._parse_where()),
1234        TokenType.GROUP_BY: lambda self: ("group", self._parse_group()),
1235        TokenType.HAVING: lambda self: ("having", self._parse_having()),
1236        TokenType.QUALIFY: lambda self: ("qualify", self._parse_qualify()),
1237        TokenType.WINDOW: lambda self: ("windows", self._parse_window_clause()),
1238        TokenType.ORDER_BY: lambda self: ("order", self._parse_order()),
1239        TokenType.LIMIT: lambda self: ("limit", self._parse_limit()),
1240        TokenType.FETCH: lambda self: ("limit", self._parse_limit()),
1241        TokenType.OFFSET: lambda self: ("offset", self._parse_offset()),
1242        TokenType.FOR: lambda self: ("locks", self._parse_locks()),
1243        TokenType.LOCK: lambda self: ("locks", self._parse_locks()),
1244        TokenType.TABLE_SAMPLE: lambda self: ("sample", self._parse_table_sample(as_modifier=True)),
1245        TokenType.USING: lambda self: ("sample", self._parse_table_sample(as_modifier=True)),
1246        TokenType.CLUSTER_BY: lambda self: (
1247            "cluster",
1248            self._parse_sort(exp.Cluster, TokenType.CLUSTER_BY),
1249        ),
1250        TokenType.DISTRIBUTE_BY: lambda self: (
1251            "distribute",
1252            self._parse_sort(exp.Distribute, TokenType.DISTRIBUTE_BY),
1253        ),
1254        TokenType.SORT_BY: lambda self: ("sort", self._parse_sort(exp.Sort, TokenType.SORT_BY)),
1255        TokenType.CONNECT_BY: lambda self: ("connect", self._parse_connect(skip_start_token=True)),
1256        TokenType.START_WITH: lambda self: ("connect", self._parse_connect()),
1257    }
1258
1259    SET_PARSERS = {
1260        "GLOBAL": lambda self: self._parse_set_item_assignment("GLOBAL"),
1261        "LOCAL": lambda self: self._parse_set_item_assignment("LOCAL"),
1262        "SESSION": lambda self: self._parse_set_item_assignment("SESSION"),
1263        "TRANSACTION": lambda self: self._parse_set_transaction(),
1264    }
1265
1266    SHOW_PARSERS: t.Dict[str, t.Callable] = {}
1267
1268    TYPE_LITERAL_PARSERS = {
1269        exp.DataType.Type.JSON: lambda self, this, _: self.expression(exp.ParseJSON, this=this),
1270    }
1271
1272    TYPE_CONVERTERS: t.Dict[exp.DataType.Type, t.Callable[[exp.DataType], exp.DataType]] = {}
1273
1274    DDL_SELECT_TOKENS = {TokenType.SELECT, TokenType.WITH, TokenType.L_PAREN}
1275
1276    PRE_VOLATILE_TOKENS = {TokenType.CREATE, TokenType.REPLACE, TokenType.UNIQUE}
1277
1278    TRANSACTION_KIND = {"DEFERRED", "IMMEDIATE", "EXCLUSIVE"}
1279    TRANSACTION_CHARACTERISTICS: OPTIONS_TYPE = {
1280        "ISOLATION": (
1281            ("LEVEL", "REPEATABLE", "READ"),
1282            ("LEVEL", "READ", "COMMITTED"),
1283            ("LEVEL", "READ", "UNCOMITTED"),
1284            ("LEVEL", "SERIALIZABLE"),
1285        ),
1286        "READ": ("WRITE", "ONLY"),
1287    }
1288
1289    CONFLICT_ACTIONS: OPTIONS_TYPE = dict.fromkeys(
1290        ("ABORT", "FAIL", "IGNORE", "REPLACE", "ROLLBACK", "UPDATE"), tuple()
1291    )
1292    CONFLICT_ACTIONS["DO"] = ("NOTHING", "UPDATE")
1293
1294    CREATE_SEQUENCE: OPTIONS_TYPE = {
1295        "SCALE": ("EXTEND", "NOEXTEND"),
1296        "SHARD": ("EXTEND", "NOEXTEND"),
1297        "NO": ("CYCLE", "CACHE", "MAXVALUE", "MINVALUE"),
1298        **dict.fromkeys(
1299            (
1300                "SESSION",
1301                "GLOBAL",
1302                "KEEP",
1303                "NOKEEP",
1304                "ORDER",
1305                "NOORDER",
1306                "NOCACHE",
1307                "CYCLE",
1308                "NOCYCLE",
1309                "NOMINVALUE",
1310                "NOMAXVALUE",
1311                "NOSCALE",
1312                "NOSHARD",
1313            ),
1314            tuple(),
1315        ),
1316    }
1317
1318    ISOLATED_LOADING_OPTIONS: OPTIONS_TYPE = {"FOR": ("ALL", "INSERT", "NONE")}
1319
1320    USABLES: OPTIONS_TYPE = dict.fromkeys(
1321        ("ROLE", "WAREHOUSE", "DATABASE", "SCHEMA", "CATALOG"), tuple()
1322    )
1323
1324    CAST_ACTIONS: OPTIONS_TYPE = dict.fromkeys(("RENAME", "ADD"), ("FIELDS",))
1325
1326    SCHEMA_BINDING_OPTIONS: OPTIONS_TYPE = {
1327        "TYPE": ("EVOLUTION",),
1328        **dict.fromkeys(("BINDING", "COMPENSATION", "EVOLUTION"), tuple()),
1329    }
1330
1331    PROCEDURE_OPTIONS: OPTIONS_TYPE = {}
1332
1333    EXECUTE_AS_OPTIONS: OPTIONS_TYPE = dict.fromkeys(("CALLER", "SELF", "OWNER"), tuple())
1334
1335    KEY_CONSTRAINT_OPTIONS: OPTIONS_TYPE = {
1336        "NOT": ("ENFORCED",),
1337        "MATCH": (
1338            "FULL",
1339            "PARTIAL",
1340            "SIMPLE",
1341        ),
1342        "INITIALLY": ("DEFERRED", "IMMEDIATE"),
1343        "USING": (
1344            "BTREE",
1345            "HASH",
1346        ),
1347        **dict.fromkeys(("DEFERRABLE", "NORELY", "RELY"), tuple()),
1348    }
1349
1350    WINDOW_EXCLUDE_OPTIONS: OPTIONS_TYPE = {
1351        "NO": ("OTHERS",),
1352        "CURRENT": ("ROW",),
1353        **dict.fromkeys(("GROUP", "TIES"), tuple()),
1354    }
1355
1356    INSERT_ALTERNATIVES = {"ABORT", "FAIL", "IGNORE", "REPLACE", "ROLLBACK"}
1357
1358    CLONE_KEYWORDS = {"CLONE", "COPY"}
1359    HISTORICAL_DATA_PREFIX = {"AT", "BEFORE", "END"}
1360    HISTORICAL_DATA_KIND = {"TIMESTAMP", "OFFSET", "STATEMENT", "STREAM"}
1361
1362    OPCLASS_FOLLOW_KEYWORDS = {"ASC", "DESC", "NULLS", "WITH"}
1363
1364    OPTYPE_FOLLOW_TOKENS = {TokenType.COMMA, TokenType.R_PAREN}
1365
1366    TABLE_INDEX_HINT_TOKENS = {TokenType.FORCE, TokenType.IGNORE, TokenType.USE}
1367
1368    VIEW_ATTRIBUTES = {"ENCRYPTION", "SCHEMABINDING", "VIEW_METADATA"}
1369
1370    WINDOW_ALIAS_TOKENS = ID_VAR_TOKENS - {TokenType.ROWS}
1371    WINDOW_BEFORE_PAREN_TOKENS = {TokenType.OVER}
1372    WINDOW_SIDES = {"FOLLOWING", "PRECEDING"}
1373
1374    JSON_KEY_VALUE_SEPARATOR_TOKENS = {TokenType.COLON, TokenType.COMMA, TokenType.IS}
1375
1376    FETCH_TOKENS = ID_VAR_TOKENS - {TokenType.ROW, TokenType.ROWS, TokenType.PERCENT}
1377
1378    ADD_CONSTRAINT_TOKENS = {
1379        TokenType.CONSTRAINT,
1380        TokenType.FOREIGN_KEY,
1381        TokenType.INDEX,
1382        TokenType.KEY,
1383        TokenType.PRIMARY_KEY,
1384        TokenType.UNIQUE,
1385    }
1386
1387    DISTINCT_TOKENS = {TokenType.DISTINCT}
1388
1389    NULL_TOKENS = {TokenType.NULL}
1390
1391    UNNEST_OFFSET_ALIAS_TOKENS = TABLE_ALIAS_TOKENS - SET_OPERATIONS
1392
1393    SELECT_START_TOKENS = {TokenType.L_PAREN, TokenType.WITH, TokenType.SELECT}
1394
1395    COPY_INTO_VARLEN_OPTIONS = {"FILE_FORMAT", "COPY_OPTIONS", "FORMAT_OPTIONS", "CREDENTIAL"}
1396
1397    IS_JSON_PREDICATE_KIND = {"VALUE", "SCALAR", "ARRAY", "OBJECT"}
1398
1399    ODBC_DATETIME_LITERALS = {
1400        "d": exp.Date,
1401        "t": exp.Time,
1402        "ts": exp.Timestamp,
1403    }
1404
1405    ON_CONDITION_TOKENS = {"ERROR", "NULL", "TRUE", "FALSE", "EMPTY"}
1406
1407    PRIVILEGE_FOLLOW_TOKENS = {TokenType.ON, TokenType.COMMA, TokenType.L_PAREN}
1408
1409    # The style options for the DESCRIBE statement
1410    DESCRIBE_STYLES = {"ANALYZE", "EXTENDED", "FORMATTED", "HISTORY"}
1411
1412    # The style options for the ANALYZE statement
1413    ANALYZE_STYLES = {
1414        "BUFFER_USAGE_LIMIT",
1415        "FULL",
1416        "LOCAL",
1417        "NO_WRITE_TO_BINLOG",
1418        "SAMPLE",
1419        "SKIP_LOCKED",
1420        "VERBOSE",
1421    }
1422
1423    ANALYZE_EXPRESSION_PARSERS = {
1424        "ALL": lambda self: self._parse_analyze_columns(),
1425        "COMPUTE": lambda self: self._parse_analyze_statistics(),
1426        "DELETE": lambda self: self._parse_analyze_delete(),
1427        "DROP": lambda self: self._parse_analyze_histogram(),
1428        "ESTIMATE": lambda self: self._parse_analyze_statistics(),
1429        "LIST": lambda self: self._parse_analyze_list(),
1430        "PREDICATE": lambda self: self._parse_analyze_columns(),
1431        "UPDATE": lambda self: self._parse_analyze_histogram(),
1432        "VALIDATE": lambda self: self._parse_analyze_validate(),
1433    }
1434
1435    PARTITION_KEYWORDS = {"PARTITION", "SUBPARTITION"}
1436
1437    AMBIGUOUS_ALIAS_TOKENS = (TokenType.LIMIT, TokenType.OFFSET)
1438
1439    OPERATION_MODIFIERS: t.Set[str] = set()
1440
1441    RECURSIVE_CTE_SEARCH_KIND = {"BREADTH", "DEPTH", "CYCLE"}
1442
1443    MODIFIABLES = (exp.Query, exp.Table, exp.TableFromRows)
1444
1445    STRICT_CAST = True
1446
1447    PREFIXED_PIVOT_COLUMNS = False
1448    IDENTIFY_PIVOT_STRINGS = False
1449
1450    LOG_DEFAULTS_TO_LN = False
1451
1452    # Whether ADD is present for each column added by ALTER TABLE
1453    ALTER_TABLE_ADD_REQUIRED_FOR_EACH_COLUMN = True
1454
1455    # Whether the table sample clause expects CSV syntax
1456    TABLESAMPLE_CSV = False
1457
1458    # The default method used for table sampling
1459    DEFAULT_SAMPLING_METHOD: t.Optional[str] = None
1460
1461    # Whether the SET command needs a delimiter (e.g. "=") for assignments
1462    SET_REQUIRES_ASSIGNMENT_DELIMITER = True
1463
1464    # Whether the TRIM function expects the characters to trim as its first argument
1465    TRIM_PATTERN_FIRST = False
1466
1467    # Whether string aliases are supported `SELECT COUNT(*) 'count'`
1468    STRING_ALIASES = False
1469
1470    # Whether query modifiers such as LIMIT are attached to the UNION node (vs its right operand)
1471    MODIFIERS_ATTACHED_TO_SET_OP = True
1472    SET_OP_MODIFIERS = {"order", "limit", "offset"}
1473
1474    # Whether to parse IF statements that aren't followed by a left parenthesis as commands
1475    NO_PAREN_IF_COMMANDS = True
1476
1477    # Whether the -> and ->> operators expect documents of type JSON (e.g. Postgres)
1478    JSON_ARROWS_REQUIRE_JSON_TYPE = False
1479
1480    # Whether the `:` operator is used to extract a value from a VARIANT column
1481    COLON_IS_VARIANT_EXTRACT = False
1482
1483    # Whether or not a VALUES keyword needs to be followed by '(' to form a VALUES clause.
1484    # If this is True and '(' is not found, the keyword will be treated as an identifier
1485    VALUES_FOLLOWED_BY_PAREN = True
1486
1487    # Whether implicit unnesting is supported, e.g. SELECT 1 FROM y.z AS z, z.a (Redshift)
1488    SUPPORTS_IMPLICIT_UNNEST = False
1489
1490    # Whether or not interval spans are supported, INTERVAL 1 YEAR TO MONTHS
1491    INTERVAL_SPANS = True
1492
1493    # Whether a PARTITION clause can follow a table reference
1494    SUPPORTS_PARTITION_SELECTION = False
1495
1496    # Whether the `name AS expr` schema/column constraint requires parentheses around `expr`
1497    WRAPPED_TRANSFORM_COLUMN_CONSTRAINT = True
1498
1499    # Whether the 'AS' keyword is optional in the CTE definition syntax
1500    OPTIONAL_ALIAS_TOKEN_CTE = True
1501
1502    __slots__ = (
1503        "error_level",
1504        "error_message_context",
1505        "max_errors",
1506        "dialect",
1507        "sql",
1508        "errors",
1509        "_tokens",
1510        "_index",
1511        "_curr",
1512        "_next",
1513        "_prev",
1514        "_prev_comments",
1515    )
1516
1517    # Autofilled
1518    SHOW_TRIE: t.Dict = {}
1519    SET_TRIE: t.Dict = {}
1520
1521    def __init__(
1522        self,
1523        error_level: t.Optional[ErrorLevel] = None,
1524        error_message_context: int = 100,
1525        max_errors: int = 3,
1526        dialect: DialectType = None,
1527    ):
1528        from sqlglot.dialects import Dialect
1529
1530        self.error_level = error_level or ErrorLevel.IMMEDIATE
1531        self.error_message_context = error_message_context
1532        self.max_errors = max_errors
1533        self.dialect = Dialect.get_or_raise(dialect)
1534        self.reset()
1535
1536    def reset(self):
1537        self.sql = ""
1538        self.errors = []
1539        self._tokens = []
1540        self._index = 0
1541        self._curr = None
1542        self._next = None
1543        self._prev = None
1544        self._prev_comments = None
1545
1546    def parse(
1547        self, raw_tokens: t.List[Token], sql: t.Optional[str] = None
1548    ) -> t.List[t.Optional[exp.Expression]]:
1549        """
1550        Parses a list of tokens and returns a list of syntax trees, one tree
1551        per parsed SQL statement.
1552
1553        Args:
1554            raw_tokens: The list of tokens.
1555            sql: The original SQL string, used to produce helpful debug messages.
1556
1557        Returns:
1558            The list of the produced syntax trees.
1559        """
1560        return self._parse(
1561            parse_method=self.__class__._parse_statement, raw_tokens=raw_tokens, sql=sql
1562        )
1563
1564    def parse_into(
1565        self,
1566        expression_types: exp.IntoType,
1567        raw_tokens: t.List[Token],
1568        sql: t.Optional[str] = None,
1569    ) -> t.List[t.Optional[exp.Expression]]:
1570        """
1571        Parses a list of tokens into a given Expression type. If a collection of Expression
1572        types is given instead, this method will try to parse the token list into each one
1573        of them, stopping at the first for which the parsing succeeds.
1574
1575        Args:
1576            expression_types: The expression type(s) to try and parse the token list into.
1577            raw_tokens: The list of tokens.
1578            sql: The original SQL string, used to produce helpful debug messages.
1579
1580        Returns:
1581            The target Expression.
1582        """
1583        errors = []
1584        for expression_type in ensure_list(expression_types):
1585            parser = self.EXPRESSION_PARSERS.get(expression_type)
1586            if not parser:
1587                raise TypeError(f"No parser registered for {expression_type}")
1588
1589            try:
1590                return self._parse(parser, raw_tokens, sql)
1591            except ParseError as e:
1592                e.errors[0]["into_expression"] = expression_type
1593                errors.append(e)
1594
1595        raise ParseError(
1596            f"Failed to parse '{sql or raw_tokens}' into {expression_types}",
1597            errors=merge_errors(errors),
1598        ) from errors[-1]
1599
1600    def _parse(
1601        self,
1602        parse_method: t.Callable[[Parser], t.Optional[exp.Expression]],
1603        raw_tokens: t.List[Token],
1604        sql: t.Optional[str] = None,
1605    ) -> t.List[t.Optional[exp.Expression]]:
1606        self.reset()
1607        self.sql = sql or ""
1608
1609        total = len(raw_tokens)
1610        chunks: t.List[t.List[Token]] = [[]]
1611
1612        for i, token in enumerate(raw_tokens):
1613            if token.token_type == TokenType.SEMICOLON:
1614                if token.comments:
1615                    chunks.append([token])
1616
1617                if i < total - 1:
1618                    chunks.append([])
1619            else:
1620                chunks[-1].append(token)
1621
1622        expressions = []
1623
1624        for tokens in chunks:
1625            self._index = -1
1626            self._tokens = tokens
1627            self._advance()
1628
1629            expressions.append(parse_method(self))
1630
1631            if self._index < len(self._tokens):
1632                self.raise_error("Invalid expression / Unexpected token")
1633
1634            self.check_errors()
1635
1636        return expressions
1637
1638    def check_errors(self) -> None:
1639        """Logs or raises any found errors, depending on the chosen error level setting."""
1640        if self.error_level == ErrorLevel.WARN:
1641            for error in self.errors:
1642                logger.error(str(error))
1643        elif self.error_level == ErrorLevel.RAISE and self.errors:
1644            raise ParseError(
1645                concat_messages(self.errors, self.max_errors),
1646                errors=merge_errors(self.errors),
1647            )
1648
1649    def raise_error(self, message: str, token: t.Optional[Token] = None) -> None:
1650        """
1651        Appends an error in the list of recorded errors or raises it, depending on the chosen
1652        error level setting.
1653        """
1654        token = token or self._curr or self._prev or Token.string("")
1655        start = token.start
1656        end = token.end + 1
1657        start_context = self.sql[max(start - self.error_message_context, 0) : start]
1658        highlight = self.sql[start:end]
1659        end_context = self.sql[end : end + self.error_message_context]
1660
1661        error = ParseError.new(
1662            f"{message}. Line {token.line}, Col: {token.col}.\n"
1663            f"  {start_context}\033[4m{highlight}\033[0m{end_context}",
1664            description=message,
1665            line=token.line,
1666            col=token.col,
1667            start_context=start_context,
1668            highlight=highlight,
1669            end_context=end_context,
1670        )
1671
1672        if self.error_level == ErrorLevel.IMMEDIATE:
1673            raise error
1674
1675        self.errors.append(error)
1676
1677    def expression(
1678        self, exp_class: t.Type[E], comments: t.Optional[t.List[str]] = None, **kwargs
1679    ) -> E:
1680        """
1681        Creates a new, validated Expression.
1682
1683        Args:
1684            exp_class: The expression class to instantiate.
1685            comments: An optional list of comments to attach to the expression.
1686            kwargs: The arguments to set for the expression along with their respective values.
1687
1688        Returns:
1689            The target expression.
1690        """
1691        instance = exp_class(**kwargs)
1692        instance.add_comments(comments) if comments else self._add_comments(instance)
1693        return self.validate_expression(instance)
1694
1695    def _add_comments(self, expression: t.Optional[exp.Expression]) -> None:
1696        if expression and self._prev_comments:
1697            expression.add_comments(self._prev_comments)
1698            self._prev_comments = None
1699
1700    def validate_expression(self, expression: E, args: t.Optional[t.List] = None) -> E:
1701        """
1702        Validates an Expression, making sure that all its mandatory arguments are set.
1703
1704        Args:
1705            expression: The expression to validate.
1706            args: An optional list of items that was used to instantiate the expression, if it's a Func.
1707
1708        Returns:
1709            The validated expression.
1710        """
1711        if self.error_level != ErrorLevel.IGNORE:
1712            for error_message in expression.error_messages(args):
1713                self.raise_error(error_message)
1714
1715        return expression
1716
1717    def _find_sql(self, start: Token, end: Token) -> str:
1718        return self.sql[start.start : end.end + 1]
1719
1720    def _is_connected(self) -> bool:
1721        return self._prev and self._curr and self._prev.end + 1 == self._curr.start
1722
1723    def _advance(self, times: int = 1) -> None:
1724        self._index += times
1725        self._curr = seq_get(self._tokens, self._index)
1726        self._next = seq_get(self._tokens, self._index + 1)
1727
1728        if self._index > 0:
1729            self._prev = self._tokens[self._index - 1]
1730            self._prev_comments = self._prev.comments
1731        else:
1732            self._prev = None
1733            self._prev_comments = None
1734
1735    def _retreat(self, index: int) -> None:
1736        if index != self._index:
1737            self._advance(index - self._index)
1738
1739    def _warn_unsupported(self) -> None:
1740        if len(self._tokens) <= 1:
1741            return
1742
1743        # We use _find_sql because self.sql may comprise multiple chunks, and we're only
1744        # interested in emitting a warning for the one being currently processed.
1745        sql = self._find_sql(self._tokens[0], self._tokens[-1])[: self.error_message_context]
1746
1747        logger.warning(
1748            f"'{sql}' contains unsupported syntax. Falling back to parsing as a 'Command'."
1749        )
1750
1751    def _parse_command(self) -> exp.Command:
1752        self._warn_unsupported()
1753        return self.expression(
1754            exp.Command,
1755            comments=self._prev_comments,
1756            this=self._prev.text.upper(),
1757            expression=self._parse_string(),
1758        )
1759
1760    def _try_parse(self, parse_method: t.Callable[[], T], retreat: bool = False) -> t.Optional[T]:
1761        """
1762        Attemps to backtrack if a parse function that contains a try/catch internally raises an error.
1763        This behavior can be different depending on the uset-set ErrorLevel, so _try_parse aims to
1764        solve this by setting & resetting the parser state accordingly
1765        """
1766        index = self._index
1767        error_level = self.error_level
1768
1769        self.error_level = ErrorLevel.IMMEDIATE
1770        try:
1771            this = parse_method()
1772        except ParseError:
1773            this = None
1774        finally:
1775            if not this or retreat:
1776                self._retreat(index)
1777            self.error_level = error_level
1778
1779        return this
1780
1781    def _parse_comment(self, allow_exists: bool = True) -> exp.Expression:
1782        start = self._prev
1783        exists = self._parse_exists() if allow_exists else None
1784
1785        self._match(TokenType.ON)
1786
1787        materialized = self._match_text_seq("MATERIALIZED")
1788        kind = self._match_set(self.CREATABLES) and self._prev
1789        if not kind:
1790            return self._parse_as_command(start)
1791
1792        if kind.token_type in (TokenType.FUNCTION, TokenType.PROCEDURE):
1793            this = self._parse_user_defined_function(kind=kind.token_type)
1794        elif kind.token_type == TokenType.TABLE:
1795            this = self._parse_table(alias_tokens=self.COMMENT_TABLE_ALIAS_TOKENS)
1796        elif kind.token_type == TokenType.COLUMN:
1797            this = self._parse_column()
1798        else:
1799            this = self._parse_id_var()
1800
1801        self._match(TokenType.IS)
1802
1803        return self.expression(
1804            exp.Comment,
1805            this=this,
1806            kind=kind.text,
1807            expression=self._parse_string(),
1808            exists=exists,
1809            materialized=materialized,
1810        )
1811
1812    def _parse_to_table(
1813        self,
1814    ) -> exp.ToTableProperty:
1815        table = self._parse_table_parts(schema=True)
1816        return self.expression(exp.ToTableProperty, this=table)
1817
1818    # https://clickhouse.com/docs/en/engines/table-engines/mergetree-family/mergetree#mergetree-table-ttl
1819    def _parse_ttl(self) -> exp.Expression:
1820        def _parse_ttl_action() -> t.Optional[exp.Expression]:
1821            this = self._parse_bitwise()
1822
1823            if self._match_text_seq("DELETE"):
1824                return self.expression(exp.MergeTreeTTLAction, this=this, delete=True)
1825            if self._match_text_seq("RECOMPRESS"):
1826                return self.expression(
1827                    exp.MergeTreeTTLAction, this=this, recompress=self._parse_bitwise()
1828                )
1829            if self._match_text_seq("TO", "DISK"):
1830                return self.expression(
1831                    exp.MergeTreeTTLAction, this=this, to_disk=self._parse_string()
1832                )
1833            if self._match_text_seq("TO", "VOLUME"):
1834                return self.expression(
1835                    exp.MergeTreeTTLAction, this=this, to_volume=self._parse_string()
1836                )
1837
1838            return this
1839
1840        expressions = self._parse_csv(_parse_ttl_action)
1841        where = self._parse_where()
1842        group = self._parse_group()
1843
1844        aggregates = None
1845        if group and self._match(TokenType.SET):
1846            aggregates = self._parse_csv(self._parse_set_item)
1847
1848        return self.expression(
1849            exp.MergeTreeTTL,
1850            expressions=expressions,
1851            where=where,
1852            group=group,
1853            aggregates=aggregates,
1854        )
1855
1856    def _parse_statement(self) -> t.Optional[exp.Expression]:
1857        if self._curr is None:
1858            return None
1859
1860        if self._match_set(self.STATEMENT_PARSERS):
1861            comments = self._prev_comments
1862            stmt = self.STATEMENT_PARSERS[self._prev.token_type](self)
1863            stmt.add_comments(comments, prepend=True)
1864            return stmt
1865
1866        if self._match_set(self.dialect.tokenizer.COMMANDS):
1867            return self._parse_command()
1868
1869        expression = self._parse_expression()
1870        expression = self._parse_set_operations(expression) if expression else self._parse_select()
1871        return self._parse_query_modifiers(expression)
1872
1873    def _parse_drop(self, exists: bool = False) -> exp.Drop | exp.Command:
1874        start = self._prev
1875        temporary = self._match(TokenType.TEMPORARY)
1876        materialized = self._match_text_seq("MATERIALIZED")
1877
1878        kind = self._match_set(self.CREATABLES) and self._prev.text.upper()
1879        if not kind:
1880            return self._parse_as_command(start)
1881
1882        concurrently = self._match_text_seq("CONCURRENTLY")
1883        if_exists = exists or self._parse_exists()
1884
1885        if kind == "COLUMN":
1886            this = self._parse_column()
1887        else:
1888            this = self._parse_table_parts(
1889                schema=True, is_db_reference=self._prev.token_type == TokenType.SCHEMA
1890            )
1891
1892        cluster = self._parse_on_property() if self._match(TokenType.ON) else None
1893
1894        if self._match(TokenType.L_PAREN, advance=False):
1895            expressions = self._parse_wrapped_csv(self._parse_types)
1896        else:
1897            expressions = None
1898
1899        return self.expression(
1900            exp.Drop,
1901            exists=if_exists,
1902            this=this,
1903            expressions=expressions,
1904            kind=self.dialect.CREATABLE_KIND_MAPPING.get(kind) or kind,
1905            temporary=temporary,
1906            materialized=materialized,
1907            cascade=self._match_text_seq("CASCADE"),
1908            constraints=self._match_text_seq("CONSTRAINTS"),
1909            purge=self._match_text_seq("PURGE"),
1910            cluster=cluster,
1911            concurrently=concurrently,
1912        )
1913
1914    def _parse_exists(self, not_: bool = False) -> t.Optional[bool]:
1915        return (
1916            self._match_text_seq("IF")
1917            and (not not_ or self._match(TokenType.NOT))
1918            and self._match(TokenType.EXISTS)
1919        )
1920
1921    def _parse_create(self) -> exp.Create | exp.Command:
1922        # Note: this can't be None because we've matched a statement parser
1923        start = self._prev
1924
1925        replace = (
1926            start.token_type == TokenType.REPLACE
1927            or self._match_pair(TokenType.OR, TokenType.REPLACE)
1928            or self._match_pair(TokenType.OR, TokenType.ALTER)
1929        )
1930        refresh = self._match_pair(TokenType.OR, TokenType.REFRESH)
1931
1932        unique = self._match(TokenType.UNIQUE)
1933
1934        if self._match_text_seq("CLUSTERED", "COLUMNSTORE"):
1935            clustered = True
1936        elif self._match_text_seq("NONCLUSTERED", "COLUMNSTORE") or self._match_text_seq(
1937            "COLUMNSTORE"
1938        ):
1939            clustered = False
1940        else:
1941            clustered = None
1942
1943        if self._match_pair(TokenType.TABLE, TokenType.FUNCTION, advance=False):
1944            self._advance()
1945
1946        properties = None
1947        create_token = self._match_set(self.CREATABLES) and self._prev
1948
1949        if not create_token:
1950            # exp.Properties.Location.POST_CREATE
1951            properties = self._parse_properties()
1952            create_token = self._match_set(self.CREATABLES) and self._prev
1953
1954            if not properties or not create_token:
1955                return self._parse_as_command(start)
1956
1957        concurrently = self._match_text_seq("CONCURRENTLY")
1958        exists = self._parse_exists(not_=True)
1959        this = None
1960        expression: t.Optional[exp.Expression] = None
1961        indexes = None
1962        no_schema_binding = None
1963        begin = None
1964        end = None
1965        clone = None
1966
1967        def extend_props(temp_props: t.Optional[exp.Properties]) -> None:
1968            nonlocal properties
1969            if properties and temp_props:
1970                properties.expressions.extend(temp_props.expressions)
1971            elif temp_props:
1972                properties = temp_props
1973
1974        if create_token.token_type in (TokenType.FUNCTION, TokenType.PROCEDURE):
1975            this = self._parse_user_defined_function(kind=create_token.token_type)
1976
1977            # exp.Properties.Location.POST_SCHEMA ("schema" here is the UDF's type signature)
1978            extend_props(self._parse_properties())
1979
1980            expression = self._match(TokenType.ALIAS) and self._parse_heredoc()
1981            extend_props(self._parse_properties())
1982
1983            if not expression:
1984                if self._match(TokenType.COMMAND):
1985                    expression = self._parse_as_command(self._prev)
1986                else:
1987                    begin = self._match(TokenType.BEGIN)
1988                    return_ = self._match_text_seq("RETURN")
1989
1990                    if self._match(TokenType.STRING, advance=False):
1991                        # Takes care of BigQuery's JavaScript UDF definitions that end in an OPTIONS property
1992                        # # https://cloud.google.com/bigquery/docs/reference/standard-sql/data-definition-language#create_function_statement
1993                        expression = self._parse_string()
1994                        extend_props(self._parse_properties())
1995                    else:
1996                        expression = self._parse_user_defined_function_expression()
1997
1998                    end = self._match_text_seq("END")
1999
2000                    if return_:
2001                        expression = self.expression(exp.Return, this=expression)
2002        elif create_token.token_type == TokenType.INDEX:
2003            # Postgres allows anonymous indexes, eg. CREATE INDEX IF NOT EXISTS ON t(c)
2004            if not self._match(TokenType.ON):
2005                index = self._parse_id_var()
2006                anonymous = False
2007            else:
2008                index = None
2009                anonymous = True
2010
2011            this = self._parse_index(index=index, anonymous=anonymous)
2012        elif create_token.token_type in self.DB_CREATABLES:
2013            table_parts = self._parse_table_parts(
2014                schema=True, is_db_reference=create_token.token_type == TokenType.SCHEMA
2015            )
2016
2017            # exp.Properties.Location.POST_NAME
2018            self._match(TokenType.COMMA)
2019            extend_props(self._parse_properties(before=True))
2020
2021            this = self._parse_schema(this=table_parts)
2022
2023            # exp.Properties.Location.POST_SCHEMA and POST_WITH
2024            extend_props(self._parse_properties())
2025
2026            has_alias = self._match(TokenType.ALIAS)
2027            if not self._match_set(self.DDL_SELECT_TOKENS, advance=False):
2028                # exp.Properties.Location.POST_ALIAS
2029                extend_props(self._parse_properties())
2030
2031            if create_token.token_type == TokenType.SEQUENCE:
2032                expression = self._parse_types()
2033                extend_props(self._parse_properties())
2034            else:
2035                expression = self._parse_ddl_select()
2036
2037                # Some dialects also support using a table as an alias instead of a SELECT.
2038                # Here we fallback to this as an alternative.
2039                if not expression and has_alias:
2040                    expression = self._try_parse(self._parse_table_parts)
2041
2042            if create_token.token_type == TokenType.TABLE:
2043                # exp.Properties.Location.POST_EXPRESSION
2044                extend_props(self._parse_properties())
2045
2046                indexes = []
2047                while True:
2048                    index = self._parse_index()
2049
2050                    # exp.Properties.Location.POST_INDEX
2051                    extend_props(self._parse_properties())
2052                    if not index:
2053                        break
2054                    else:
2055                        self._match(TokenType.COMMA)
2056                        indexes.append(index)
2057            elif create_token.token_type == TokenType.VIEW:
2058                if self._match_text_seq("WITH", "NO", "SCHEMA", "BINDING"):
2059                    no_schema_binding = True
2060            elif create_token.token_type in (TokenType.SINK, TokenType.SOURCE):
2061                extend_props(self._parse_properties())
2062
2063            shallow = self._match_text_seq("SHALLOW")
2064
2065            if self._match_texts(self.CLONE_KEYWORDS):
2066                copy = self._prev.text.lower() == "copy"
2067                clone = self.expression(
2068                    exp.Clone, this=self._parse_table(schema=True), shallow=shallow, copy=copy
2069                )
2070
2071        if self._curr and not self._match_set((TokenType.R_PAREN, TokenType.COMMA), advance=False):
2072            return self._parse_as_command(start)
2073
2074        create_kind_text = create_token.text.upper()
2075        return self.expression(
2076            exp.Create,
2077            this=this,
2078            kind=self.dialect.CREATABLE_KIND_MAPPING.get(create_kind_text) or create_kind_text,
2079            replace=replace,
2080            refresh=refresh,
2081            unique=unique,
2082            expression=expression,
2083            exists=exists,
2084            properties=properties,
2085            indexes=indexes,
2086            no_schema_binding=no_schema_binding,
2087            begin=begin,
2088            end=end,
2089            clone=clone,
2090            concurrently=concurrently,
2091            clustered=clustered,
2092        )
2093
2094    def _parse_sequence_properties(self) -> t.Optional[exp.SequenceProperties]:
2095        seq = exp.SequenceProperties()
2096
2097        options = []
2098        index = self._index
2099
2100        while self._curr:
2101            self._match(TokenType.COMMA)
2102            if self._match_text_seq("INCREMENT"):
2103                self._match_text_seq("BY")
2104                self._match_text_seq("=")
2105                seq.set("increment", self._parse_term())
2106            elif self._match_text_seq("MINVALUE"):
2107                seq.set("minvalue", self._parse_term())
2108            elif self._match_text_seq("MAXVALUE"):
2109                seq.set("maxvalue", self._parse_term())
2110            elif self._match(TokenType.START_WITH) or self._match_text_seq("START"):
2111                self._match_text_seq("=")
2112                seq.set("start", self._parse_term())
2113            elif self._match_text_seq("CACHE"):
2114                # T-SQL allows empty CACHE which is initialized dynamically
2115                seq.set("cache", self._parse_number() or True)
2116            elif self._match_text_seq("OWNED", "BY"):
2117                # "OWNED BY NONE" is the default
2118                seq.set("owned", None if self._match_text_seq("NONE") else self._parse_column())
2119            else:
2120                opt = self._parse_var_from_options(self.CREATE_SEQUENCE, raise_unmatched=False)
2121                if opt:
2122                    options.append(opt)
2123                else:
2124                    break
2125
2126        seq.set("options", options if options else None)
2127        return None if self._index == index else seq
2128
2129    def _parse_property_before(self) -> t.Optional[exp.Expression]:
2130        # only used for teradata currently
2131        self._match(TokenType.COMMA)
2132
2133        kwargs = {
2134            "no": self._match_text_seq("NO"),
2135            "dual": self._match_text_seq("DUAL"),
2136            "before": self._match_text_seq("BEFORE"),
2137            "default": self._match_text_seq("DEFAULT"),
2138            "local": (self._match_text_seq("LOCAL") and "LOCAL")
2139            or (self._match_text_seq("NOT", "LOCAL") and "NOT LOCAL"),
2140            "after": self._match_text_seq("AFTER"),
2141            "minimum": self._match_texts(("MIN", "MINIMUM")),
2142            "maximum": self._match_texts(("MAX", "MAXIMUM")),
2143        }
2144
2145        if self._match_texts(self.PROPERTY_PARSERS):
2146            parser = self.PROPERTY_PARSERS[self._prev.text.upper()]
2147            try:
2148                return parser(self, **{k: v for k, v in kwargs.items() if v})
2149            except TypeError:
2150                self.raise_error(f"Cannot parse property '{self._prev.text}'")
2151
2152        return None
2153
2154    def _parse_wrapped_properties(self) -> t.List[exp.Expression]:
2155        return self._parse_wrapped_csv(self._parse_property)
2156
2157    def _parse_property(self) -> t.Optional[exp.Expression]:
2158        if self._match_texts(self.PROPERTY_PARSERS):
2159            return self.PROPERTY_PARSERS[self._prev.text.upper()](self)
2160
2161        if self._match(TokenType.DEFAULT) and self._match_texts(self.PROPERTY_PARSERS):
2162            return self.PROPERTY_PARSERS[self._prev.text.upper()](self, default=True)
2163
2164        if self._match_text_seq("COMPOUND", "SORTKEY"):
2165            return self._parse_sortkey(compound=True)
2166
2167        if self._match_text_seq("SQL", "SECURITY"):
2168            return self.expression(exp.SqlSecurityProperty, definer=self._match_text_seq("DEFINER"))
2169
2170        index = self._index
2171        key = self._parse_column()
2172
2173        if not self._match(TokenType.EQ):
2174            self._retreat(index)
2175            return self._parse_sequence_properties()
2176
2177        # Transform the key to exp.Dot if it's dotted identifiers wrapped in exp.Column or to exp.Var otherwise
2178        if isinstance(key, exp.Column):
2179            key = key.to_dot() if len(key.parts) > 1 else exp.var(key.name)
2180
2181        value = self._parse_bitwise() or self._parse_var(any_token=True)
2182
2183        # Transform the value to exp.Var if it was parsed as exp.Column(exp.Identifier())
2184        if isinstance(value, exp.Column):
2185            value = exp.var(value.name)
2186
2187        return self.expression(exp.Property, this=key, value=value)
2188
2189    def _parse_stored(self) -> t.Union[exp.FileFormatProperty, exp.StorageHandlerProperty]:
2190        if self._match_text_seq("BY"):
2191            return self.expression(exp.StorageHandlerProperty, this=self._parse_var_or_string())
2192
2193        self._match(TokenType.ALIAS)
2194        input_format = self._parse_string() if self._match_text_seq("INPUTFORMAT") else None
2195        output_format = self._parse_string() if self._match_text_seq("OUTPUTFORMAT") else None
2196
2197        return self.expression(
2198            exp.FileFormatProperty,
2199            this=(
2200                self.expression(
2201                    exp.InputOutputFormat,
2202                    input_format=input_format,
2203                    output_format=output_format,
2204                )
2205                if input_format or output_format
2206                else self._parse_var_or_string() or self._parse_number() or self._parse_id_var()
2207            ),
2208        )
2209
2210    def _parse_unquoted_field(self) -> t.Optional[exp.Expression]:
2211        field = self._parse_field()
2212        if isinstance(field, exp.Identifier) and not field.quoted:
2213            field = exp.var(field)
2214
2215        return field
2216
2217    def _parse_property_assignment(self, exp_class: t.Type[E], **kwargs: t.Any) -> E:
2218        self._match(TokenType.EQ)
2219        self._match(TokenType.ALIAS)
2220
2221        return self.expression(exp_class, this=self._parse_unquoted_field(), **kwargs)
2222
2223    def _parse_properties(self, before: t.Optional[bool] = None) -> t.Optional[exp.Properties]:
2224        properties = []
2225        while True:
2226            if before:
2227                prop = self._parse_property_before()
2228            else:
2229                prop = self._parse_property()
2230            if not prop:
2231                break
2232            for p in ensure_list(prop):
2233                properties.append(p)
2234
2235        if properties:
2236            return self.expression(exp.Properties, expressions=properties)
2237
2238        return None
2239
2240    def _parse_fallback(self, no: bool = False) -> exp.FallbackProperty:
2241        return self.expression(
2242            exp.FallbackProperty, no=no, protection=self._match_text_seq("PROTECTION")
2243        )
2244
2245    def _parse_security(self) -> t.Optional[exp.SecurityProperty]:
2246        if self._match_texts(("NONE", "DEFINER", "INVOKER")):
2247            security_specifier = self._prev.text.upper()
2248            return self.expression(exp.SecurityProperty, this=security_specifier)
2249        return None
2250
2251    def _parse_settings_property(self) -> exp.SettingsProperty:
2252        return self.expression(
2253            exp.SettingsProperty, expressions=self._parse_csv(self._parse_assignment)
2254        )
2255
2256    def _parse_volatile_property(self) -> exp.VolatileProperty | exp.StabilityProperty:
2257        if self._index >= 2:
2258            pre_volatile_token = self._tokens[self._index - 2]
2259        else:
2260            pre_volatile_token = None
2261
2262        if pre_volatile_token and pre_volatile_token.token_type in self.PRE_VOLATILE_TOKENS:
2263            return exp.VolatileProperty()
2264
2265        return self.expression(exp.StabilityProperty, this=exp.Literal.string("VOLATILE"))
2266
2267    def _parse_retention_period(self) -> exp.Var:
2268        # Parse TSQL's HISTORY_RETENTION_PERIOD: {INFINITE | <number> DAY | DAYS | MONTH ...}
2269        number = self._parse_number()
2270        number_str = f"{number} " if number else ""
2271        unit = self._parse_var(any_token=True)
2272        return exp.var(f"{number_str}{unit}")
2273
2274    def _parse_system_versioning_property(
2275        self, with_: bool = False
2276    ) -> exp.WithSystemVersioningProperty:
2277        self._match(TokenType.EQ)
2278        prop = self.expression(
2279            exp.WithSystemVersioningProperty,
2280            **{  # type: ignore
2281                "on": True,
2282                "with": with_,
2283            },
2284        )
2285
2286        if self._match_text_seq("OFF"):
2287            prop.set("on", False)
2288            return prop
2289
2290        self._match(TokenType.ON)
2291        if self._match(TokenType.L_PAREN):
2292            while self._curr and not self._match(TokenType.R_PAREN):
2293                if self._match_text_seq("HISTORY_TABLE", "="):
2294                    prop.set("this", self._parse_table_parts())
2295                elif self._match_text_seq("DATA_CONSISTENCY_CHECK", "="):
2296                    prop.set("data_consistency", self._advance_any() and self._prev.text.upper())
2297                elif self._match_text_seq("HISTORY_RETENTION_PERIOD", "="):
2298                    prop.set("retention_period", self._parse_retention_period())
2299
2300                self._match(TokenType.COMMA)
2301
2302        return prop
2303
2304    def _parse_data_deletion_property(self) -> exp.DataDeletionProperty:
2305        self._match(TokenType.EQ)
2306        on = self._match_text_seq("ON") or not self._match_text_seq("OFF")
2307        prop = self.expression(exp.DataDeletionProperty, on=on)
2308
2309        if self._match(TokenType.L_PAREN):
2310            while self._curr and not self._match(TokenType.R_PAREN):
2311                if self._match_text_seq("FILTER_COLUMN", "="):
2312                    prop.set("filter_column", self._parse_column())
2313                elif self._match_text_seq("RETENTION_PERIOD", "="):
2314                    prop.set("retention_period", self._parse_retention_period())
2315
2316                self._match(TokenType.COMMA)
2317
2318        return prop
2319
2320    def _parse_distributed_property(self) -> exp.DistributedByProperty:
2321        kind = "HASH"
2322        expressions: t.Optional[t.List[exp.Expression]] = None
2323        if self._match_text_seq("BY", "HASH"):
2324            expressions = self._parse_wrapped_csv(self._parse_id_var)
2325        elif self._match_text_seq("BY", "RANDOM"):
2326            kind = "RANDOM"
2327
2328        # If the BUCKETS keyword is not present, the number of buckets is AUTO
2329        buckets: t.Optional[exp.Expression] = None
2330        if self._match_text_seq("BUCKETS") and not self._match_text_seq("AUTO"):
2331            buckets = self._parse_number()
2332
2333        return self.expression(
2334            exp.DistributedByProperty,
2335            expressions=expressions,
2336            kind=kind,
2337            buckets=buckets,
2338            order=self._parse_order(),
2339        )
2340
2341    def _parse_composite_key_property(self, expr_type: t.Type[E]) -> E:
2342        self._match_text_seq("KEY")
2343        expressions = self._parse_wrapped_id_vars()
2344        return self.expression(expr_type, expressions=expressions)
2345
2346    def _parse_with_property(self) -> t.Optional[exp.Expression] | t.List[exp.Expression]:
2347        if self._match_text_seq("(", "SYSTEM_VERSIONING"):
2348            prop = self._parse_system_versioning_property(with_=True)
2349            self._match_r_paren()
2350            return prop
2351
2352        if self._match(TokenType.L_PAREN, advance=False):
2353            return self._parse_wrapped_properties()
2354
2355        if self._match_text_seq("JOURNAL"):
2356            return self._parse_withjournaltable()
2357
2358        if self._match_texts(self.VIEW_ATTRIBUTES):
2359            return self.expression(exp.ViewAttributeProperty, this=self._prev.text.upper())
2360
2361        if self._match_text_seq("DATA"):
2362            return self._parse_withdata(no=False)
2363        elif self._match_text_seq("NO", "DATA"):
2364            return self._parse_withdata(no=True)
2365
2366        if self._match(TokenType.SERDE_PROPERTIES, advance=False):
2367            return self._parse_serde_properties(with_=True)
2368
2369        if self._match(TokenType.SCHEMA):
2370            return self.expression(
2371                exp.WithSchemaBindingProperty,
2372                this=self._parse_var_from_options(self.SCHEMA_BINDING_OPTIONS),
2373            )
2374
2375        if self._match_texts(self.PROCEDURE_OPTIONS, advance=False):
2376            return self.expression(
2377                exp.WithProcedureOptions, expressions=self._parse_csv(self._parse_procedure_option)
2378            )
2379
2380        if not self._next:
2381            return None
2382
2383        return self._parse_withisolatedloading()
2384
2385    def _parse_procedure_option(self) -> exp.Expression | None:
2386        if self._match_text_seq("EXECUTE", "AS"):
2387            return self.expression(
2388                exp.ExecuteAsProperty,
2389                this=self._parse_var_from_options(self.EXECUTE_AS_OPTIONS, raise_unmatched=False)
2390                or self._parse_string(),
2391            )
2392
2393        return self._parse_var_from_options(self.PROCEDURE_OPTIONS)
2394
2395    # https://dev.mysql.com/doc/refman/8.0/en/create-view.html
2396    def _parse_definer(self) -> t.Optional[exp.DefinerProperty]:
2397        self._match(TokenType.EQ)
2398
2399        user = self._parse_id_var()
2400        self._match(TokenType.PARAMETER)
2401        host = self._parse_id_var() or (self._match(TokenType.MOD) and self._prev.text)
2402
2403        if not user or not host:
2404            return None
2405
2406        return exp.DefinerProperty(this=f"{user}@{host}")
2407
2408    def _parse_withjournaltable(self) -> exp.WithJournalTableProperty:
2409        self._match(TokenType.TABLE)
2410        self._match(TokenType.EQ)
2411        return self.expression(exp.WithJournalTableProperty, this=self._parse_table_parts())
2412
2413    def _parse_log(self, no: bool = False) -> exp.LogProperty:
2414        return self.expression(exp.LogProperty, no=no)
2415
2416    def _parse_journal(self, **kwargs) -> exp.JournalProperty:
2417        return self.expression(exp.JournalProperty, **kwargs)
2418
2419    def _parse_checksum(self) -> exp.ChecksumProperty:
2420        self._match(TokenType.EQ)
2421
2422        on = None
2423        if self._match(TokenType.ON):
2424            on = True
2425        elif self._match_text_seq("OFF"):
2426            on = False
2427
2428        return self.expression(exp.ChecksumProperty, on=on, default=self._match(TokenType.DEFAULT))
2429
2430    def _parse_cluster(self, wrapped: bool = False) -> exp.Cluster:
2431        return self.expression(
2432            exp.Cluster,
2433            expressions=(
2434                self._parse_wrapped_csv(self._parse_ordered)
2435                if wrapped
2436                else self._parse_csv(self._parse_ordered)
2437            ),
2438        )
2439
2440    def _parse_clustered_by(self) -> exp.ClusteredByProperty:
2441        self._match_text_seq("BY")
2442
2443        self._match_l_paren()
2444        expressions = self._parse_csv(self._parse_column)
2445        self._match_r_paren()
2446
2447        if self._match_text_seq("SORTED", "BY"):
2448            self._match_l_paren()
2449            sorted_by = self._parse_csv(self._parse_ordered)
2450            self._match_r_paren()
2451        else:
2452            sorted_by = None
2453
2454        self._match(TokenType.INTO)
2455        buckets = self._parse_number()
2456        self._match_text_seq("BUCKETS")
2457
2458        return self.expression(
2459            exp.ClusteredByProperty,
2460            expressions=expressions,
2461            sorted_by=sorted_by,
2462            buckets=buckets,
2463        )
2464
2465    def _parse_copy_property(self) -> t.Optional[exp.CopyGrantsProperty]:
2466        if not self._match_text_seq("GRANTS"):
2467            self._retreat(self._index - 1)
2468            return None
2469
2470        return self.expression(exp.CopyGrantsProperty)
2471
2472    def _parse_freespace(self) -> exp.FreespaceProperty:
2473        self._match(TokenType.EQ)
2474        return self.expression(
2475            exp.FreespaceProperty, this=self._parse_number(), percent=self._match(TokenType.PERCENT)
2476        )
2477
2478    def _parse_mergeblockratio(
2479        self, no: bool = False, default: bool = False
2480    ) -> exp.MergeBlockRatioProperty:
2481        if self._match(TokenType.EQ):
2482            return self.expression(
2483                exp.MergeBlockRatioProperty,
2484                this=self._parse_number(),
2485                percent=self._match(TokenType.PERCENT),
2486            )
2487
2488        return self.expression(exp.MergeBlockRatioProperty, no=no, default=default)
2489
2490    def _parse_datablocksize(
2491        self,
2492        default: t.Optional[bool] = None,
2493        minimum: t.Optional[bool] = None,
2494        maximum: t.Optional[bool] = None,
2495    ) -> exp.DataBlocksizeProperty:
2496        self._match(TokenType.EQ)
2497        size = self._parse_number()
2498
2499        units = None
2500        if self._match_texts(("BYTES", "KBYTES", "KILOBYTES")):
2501            units = self._prev.text
2502
2503        return self.expression(
2504            exp.DataBlocksizeProperty,
2505            size=size,
2506            units=units,
2507            default=default,
2508            minimum=minimum,
2509            maximum=maximum,
2510        )
2511
2512    def _parse_blockcompression(self) -> exp.BlockCompressionProperty:
2513        self._match(TokenType.EQ)
2514        always = self._match_text_seq("ALWAYS")
2515        manual = self._match_text_seq("MANUAL")
2516        never = self._match_text_seq("NEVER")
2517        default = self._match_text_seq("DEFAULT")
2518
2519        autotemp = None
2520        if self._match_text_seq("AUTOTEMP"):
2521            autotemp = self._parse_schema()
2522
2523        return self.expression(
2524            exp.BlockCompressionProperty,
2525            always=always,
2526            manual=manual,
2527            never=never,
2528            default=default,
2529            autotemp=autotemp,
2530        )
2531
2532    def _parse_withisolatedloading(self) -> t.Optional[exp.IsolatedLoadingProperty]:
2533        index = self._index
2534        no = self._match_text_seq("NO")
2535        concurrent = self._match_text_seq("CONCURRENT")
2536
2537        if not self._match_text_seq("ISOLATED", "LOADING"):
2538            self._retreat(index)
2539            return None
2540
2541        target = self._parse_var_from_options(self.ISOLATED_LOADING_OPTIONS, raise_unmatched=False)
2542        return self.expression(
2543            exp.IsolatedLoadingProperty, no=no, concurrent=concurrent, target=target
2544        )
2545
2546    def _parse_locking(self) -> exp.LockingProperty:
2547        if self._match(TokenType.TABLE):
2548            kind = "TABLE"
2549        elif self._match(TokenType.VIEW):
2550            kind = "VIEW"
2551        elif self._match(TokenType.ROW):
2552            kind = "ROW"
2553        elif self._match_text_seq("DATABASE"):
2554            kind = "DATABASE"
2555        else:
2556            kind = None
2557
2558        if kind in ("DATABASE", "TABLE", "VIEW"):
2559            this = self._parse_table_parts()
2560        else:
2561            this = None
2562
2563        if self._match(TokenType.FOR):
2564            for_or_in = "FOR"
2565        elif self._match(TokenType.IN):
2566            for_or_in = "IN"
2567        else:
2568            for_or_in = None
2569
2570        if self._match_text_seq("ACCESS"):
2571            lock_type = "ACCESS"
2572        elif self._match_texts(("EXCL", "EXCLUSIVE")):
2573            lock_type = "EXCLUSIVE"
2574        elif self._match_text_seq("SHARE"):
2575            lock_type = "SHARE"
2576        elif self._match_text_seq("READ"):
2577            lock_type = "READ"
2578        elif self._match_text_seq("WRITE"):
2579            lock_type = "WRITE"
2580        elif self._match_text_seq("CHECKSUM"):
2581            lock_type = "CHECKSUM"
2582        else:
2583            lock_type = None
2584
2585        override = self._match_text_seq("OVERRIDE")
2586
2587        return self.expression(
2588            exp.LockingProperty,
2589            this=this,
2590            kind=kind,
2591            for_or_in=for_or_in,
2592            lock_type=lock_type,
2593            override=override,
2594        )
2595
2596    def _parse_partition_by(self) -> t.List[exp.Expression]:
2597        if self._match(TokenType.PARTITION_BY):
2598            return self._parse_csv(self._parse_assignment)
2599        return []
2600
2601    def _parse_partition_bound_spec(self) -> exp.PartitionBoundSpec:
2602        def _parse_partition_bound_expr() -> t.Optional[exp.Expression]:
2603            if self._match_text_seq("MINVALUE"):
2604                return exp.var("MINVALUE")
2605            if self._match_text_seq("MAXVALUE"):
2606                return exp.var("MAXVALUE")
2607            return self._parse_bitwise()
2608
2609        this: t.Optional[exp.Expression | t.List[exp.Expression]] = None
2610        expression = None
2611        from_expressions = None
2612        to_expressions = None
2613
2614        if self._match(TokenType.IN):
2615            this = self._parse_wrapped_csv(self._parse_bitwise)
2616        elif self._match(TokenType.FROM):
2617            from_expressions = self._parse_wrapped_csv(_parse_partition_bound_expr)
2618            self._match_text_seq("TO")
2619            to_expressions = self._parse_wrapped_csv(_parse_partition_bound_expr)
2620        elif self._match_text_seq("WITH", "(", "MODULUS"):
2621            this = self._parse_number()
2622            self._match_text_seq(",", "REMAINDER")
2623            expression = self._parse_number()
2624            self._match_r_paren()
2625        else:
2626            self.raise_error("Failed to parse partition bound spec.")
2627
2628        return self.expression(
2629            exp.PartitionBoundSpec,
2630            this=this,
2631            expression=expression,
2632            from_expressions=from_expressions,
2633            to_expressions=to_expressions,
2634        )
2635
2636    # https://www.postgresql.org/docs/current/sql-createtable.html
2637    def _parse_partitioned_of(self) -> t.Optional[exp.PartitionedOfProperty]:
2638        if not self._match_text_seq("OF"):
2639            self._retreat(self._index - 1)
2640            return None
2641
2642        this = self._parse_table(schema=True)
2643
2644        if self._match(TokenType.DEFAULT):
2645            expression: exp.Var | exp.PartitionBoundSpec = exp.var("DEFAULT")
2646        elif self._match_text_seq("FOR", "VALUES"):
2647            expression = self._parse_partition_bound_spec()
2648        else:
2649            self.raise_error("Expecting either DEFAULT or FOR VALUES clause.")
2650
2651        return self.expression(exp.PartitionedOfProperty, this=this, expression=expression)
2652
2653    def _parse_partitioned_by(self) -> exp.PartitionedByProperty:
2654        self._match(TokenType.EQ)
2655        return self.expression(
2656            exp.PartitionedByProperty,
2657            this=self._parse_schema() or self._parse_bracket(self._parse_field()),
2658        )
2659
2660    def _parse_withdata(self, no: bool = False) -> exp.WithDataProperty:
2661        if self._match_text_seq("AND", "STATISTICS"):
2662            statistics = True
2663        elif self._match_text_seq("AND", "NO", "STATISTICS"):
2664            statistics = False
2665        else:
2666            statistics = None
2667
2668        return self.expression(exp.WithDataProperty, no=no, statistics=statistics)
2669
2670    def _parse_contains_property(self) -> t.Optional[exp.SqlReadWriteProperty]:
2671        if self._match_text_seq("SQL"):
2672            return self.expression(exp.SqlReadWriteProperty, this="CONTAINS SQL")
2673        return None
2674
2675    def _parse_modifies_property(self) -> t.Optional[exp.SqlReadWriteProperty]:
2676        if self._match_text_seq("SQL", "DATA"):
2677            return self.expression(exp.SqlReadWriteProperty, this="MODIFIES SQL DATA")
2678        return None
2679
2680    def _parse_no_property(self) -> t.Optional[exp.Expression]:
2681        if self._match_text_seq("PRIMARY", "INDEX"):
2682            return exp.NoPrimaryIndexProperty()
2683        if self._match_text_seq("SQL"):
2684            return self.expression(exp.SqlReadWriteProperty, this="NO SQL")
2685        return None
2686
2687    def _parse_on_property(self) -> t.Optional[exp.Expression]:
2688        if self._match_text_seq("COMMIT", "PRESERVE", "ROWS"):
2689            return exp.OnCommitProperty()
2690        if self._match_text_seq("COMMIT", "DELETE", "ROWS"):
2691            return exp.OnCommitProperty(delete=True)
2692        return self.expression(exp.OnProperty, this=self._parse_schema(self._parse_id_var()))
2693
2694    def _parse_reads_property(self) -> t.Optional[exp.SqlReadWriteProperty]:
2695        if self._match_text_seq("SQL", "DATA"):
2696            return self.expression(exp.SqlReadWriteProperty, this="READS SQL DATA")
2697        return None
2698
2699    def _parse_distkey(self) -> exp.DistKeyProperty:
2700        return self.expression(exp.DistKeyProperty, this=self._parse_wrapped(self._parse_id_var))
2701
2702    def _parse_create_like(self) -> t.Optional[exp.LikeProperty]:
2703        table = self._parse_table(schema=True)
2704
2705        options = []
2706        while self._match_texts(("INCLUDING", "EXCLUDING")):
2707            this = self._prev.text.upper()
2708
2709            id_var = self._parse_id_var()
2710            if not id_var:
2711                return None
2712
2713            options.append(
2714                self.expression(exp.Property, this=this, value=exp.var(id_var.this.upper()))
2715            )
2716
2717        return self.expression(exp.LikeProperty, this=table, expressions=options)
2718
2719    def _parse_sortkey(self, compound: bool = False) -> exp.SortKeyProperty:
2720        return self.expression(
2721            exp.SortKeyProperty, this=self._parse_wrapped_id_vars(), compound=compound
2722        )
2723
2724    def _parse_character_set(self, default: bool = False) -> exp.CharacterSetProperty:
2725        self._match(TokenType.EQ)
2726        return self.expression(
2727            exp.CharacterSetProperty, this=self._parse_var_or_string(), default=default
2728        )
2729
2730    def _parse_remote_with_connection(self) -> exp.RemoteWithConnectionModelProperty:
2731        self._match_text_seq("WITH", "CONNECTION")
2732        return self.expression(
2733            exp.RemoteWithConnectionModelProperty, this=self._parse_table_parts()
2734        )
2735
2736    def _parse_returns(self) -> exp.ReturnsProperty:
2737        value: t.Optional[exp.Expression]
2738        null = None
2739        is_table = self._match(TokenType.TABLE)
2740
2741        if is_table:
2742            if self._match(TokenType.LT):
2743                value = self.expression(
2744                    exp.Schema,
2745                    this="TABLE",
2746                    expressions=self._parse_csv(self._parse_struct_types),
2747                )
2748                if not self._match(TokenType.GT):
2749                    self.raise_error("Expecting >")
2750            else:
2751                value = self._parse_schema(exp.var("TABLE"))
2752        elif self._match_text_seq("NULL", "ON", "NULL", "INPUT"):
2753            null = True
2754            value = None
2755        else:
2756            value = self._parse_types()
2757
2758        return self.expression(exp.ReturnsProperty, this=value, is_table=is_table, null=null)
2759
2760    def _parse_describe(self) -> exp.Describe:
2761        kind = self._match_set(self.CREATABLES) and self._prev.text
2762        style = self._match_texts(self.DESCRIBE_STYLES) and self._prev.text.upper()
2763        if self._match(TokenType.DOT):
2764            style = None
2765            self._retreat(self._index - 2)
2766
2767        format = self._parse_property() if self._match(TokenType.FORMAT, advance=False) else None
2768
2769        if self._match_set(self.STATEMENT_PARSERS, advance=False):
2770            this = self._parse_statement()
2771        else:
2772            this = self._parse_table(schema=True)
2773
2774        properties = self._parse_properties()
2775        expressions = properties.expressions if properties else None
2776        partition = self._parse_partition()
2777        return self.expression(
2778            exp.Describe,
2779            this=this,
2780            style=style,
2781            kind=kind,
2782            expressions=expressions,
2783            partition=partition,
2784            format=format,
2785        )
2786
2787    def _parse_multitable_inserts(self, comments: t.Optional[t.List[str]]) -> exp.MultitableInserts:
2788        kind = self._prev.text.upper()
2789        expressions = []
2790
2791        def parse_conditional_insert() -> t.Optional[exp.ConditionalInsert]:
2792            if self._match(TokenType.WHEN):
2793                expression = self._parse_disjunction()
2794                self._match(TokenType.THEN)
2795            else:
2796                expression = None
2797
2798            else_ = self._match(TokenType.ELSE)
2799
2800            if not self._match(TokenType.INTO):
2801                return None
2802
2803            return self.expression(
2804                exp.ConditionalInsert,
2805                this=self.expression(
2806                    exp.Insert,
2807                    this=self._parse_table(schema=True),
2808                    expression=self._parse_derived_table_values(),
2809                ),
2810                expression=expression,
2811                else_=else_,
2812            )
2813
2814        expression = parse_conditional_insert()
2815        while expression is not None:
2816            expressions.append(expression)
2817            expression = parse_conditional_insert()
2818
2819        return self.expression(
2820            exp.MultitableInserts,
2821            kind=kind,
2822            comments=comments,
2823            expressions=expressions,
2824            source=self._parse_table(),
2825        )
2826
2827    def _parse_insert(self) -> t.Union[exp.Insert, exp.MultitableInserts]:
2828        comments = []
2829        hint = self._parse_hint()
2830        overwrite = self._match(TokenType.OVERWRITE)
2831        ignore = self._match(TokenType.IGNORE)
2832        local = self._match_text_seq("LOCAL")
2833        alternative = None
2834        is_function = None
2835
2836        if self._match_text_seq("DIRECTORY"):
2837            this: t.Optional[exp.Expression] = self.expression(
2838                exp.Directory,
2839                this=self._parse_var_or_string(),
2840                local=local,
2841                row_format=self._parse_row_format(match_row=True),
2842            )
2843        else:
2844            if self._match_set((TokenType.FIRST, TokenType.ALL)):
2845                comments += ensure_list(self._prev_comments)
2846                return self._parse_multitable_inserts(comments)
2847
2848            if self._match(TokenType.OR):
2849                alternative = self._match_texts(self.INSERT_ALTERNATIVES) and self._prev.text
2850
2851            self._match(TokenType.INTO)
2852            comments += ensure_list(self._prev_comments)
2853            self._match(TokenType.TABLE)
2854            is_function = self._match(TokenType.FUNCTION)
2855
2856            this = (
2857                self._parse_table(schema=True, parse_partition=True)
2858                if not is_function
2859                else self._parse_function()
2860            )
2861            if isinstance(this, exp.Table) and self._match(TokenType.ALIAS, advance=False):
2862                this.set("alias", self._parse_table_alias())
2863
2864        returning = self._parse_returning()
2865
2866        return self.expression(
2867            exp.Insert,
2868            comments=comments,
2869            hint=hint,
2870            is_function=is_function,
2871            this=this,
2872            stored=self._match_text_seq("STORED") and self._parse_stored(),
2873            by_name=self._match_text_seq("BY", "NAME"),
2874            exists=self._parse_exists(),
2875            where=self._match_pair(TokenType.REPLACE, TokenType.WHERE) and self._parse_assignment(),
2876            partition=self._match(TokenType.PARTITION_BY) and self._parse_partitioned_by(),
2877            settings=self._match_text_seq("SETTINGS") and self._parse_settings_property(),
2878            expression=self._parse_derived_table_values() or self._parse_ddl_select(),
2879            conflict=self._parse_on_conflict(),
2880            returning=returning or self._parse_returning(),
2881            overwrite=overwrite,
2882            alternative=alternative,
2883            ignore=ignore,
2884            source=self._match(TokenType.TABLE) and self._parse_table(),
2885        )
2886
2887    def _parse_kill(self) -> exp.Kill:
2888        kind = exp.var(self._prev.text) if self._match_texts(("CONNECTION", "QUERY")) else None
2889
2890        return self.expression(
2891            exp.Kill,
2892            this=self._parse_primary(),
2893            kind=kind,
2894        )
2895
2896    def _parse_on_conflict(self) -> t.Optional[exp.OnConflict]:
2897        conflict = self._match_text_seq("ON", "CONFLICT")
2898        duplicate = self._match_text_seq("ON", "DUPLICATE", "KEY")
2899
2900        if not conflict and not duplicate:
2901            return None
2902
2903        conflict_keys = None
2904        constraint = None
2905
2906        if conflict:
2907            if self._match_text_seq("ON", "CONSTRAINT"):
2908                constraint = self._parse_id_var()
2909            elif self._match(TokenType.L_PAREN):
2910                conflict_keys = self._parse_csv(self._parse_id_var)
2911                self._match_r_paren()
2912
2913        action = self._parse_var_from_options(self.CONFLICT_ACTIONS)
2914        if self._prev.token_type == TokenType.UPDATE:
2915            self._match(TokenType.SET)
2916            expressions = self._parse_csv(self._parse_equality)
2917        else:
2918            expressions = None
2919
2920        return self.expression(
2921            exp.OnConflict,
2922            duplicate=duplicate,
2923            expressions=expressions,
2924            action=action,
2925            conflict_keys=conflict_keys,
2926            constraint=constraint,
2927            where=self._parse_where(),
2928        )
2929
2930    def _parse_returning(self) -> t.Optional[exp.Returning]:
2931        if not self._match(TokenType.RETURNING):
2932            return None
2933        return self.expression(
2934            exp.Returning,
2935            expressions=self._parse_csv(self._parse_expression),
2936            into=self._match(TokenType.INTO) and self._parse_table_part(),
2937        )
2938
2939    def _parse_row(self) -> t.Optional[exp.RowFormatSerdeProperty | exp.RowFormatDelimitedProperty]:
2940        if not self._match(TokenType.FORMAT):
2941            return None
2942        return self._parse_row_format()
2943
2944    def _parse_serde_properties(self, with_: bool = False) -> t.Optional[exp.SerdeProperties]:
2945        index = self._index
2946        with_ = with_ or self._match_text_seq("WITH")
2947
2948        if not self._match(TokenType.SERDE_PROPERTIES):
2949            self._retreat(index)
2950            return None
2951        return self.expression(
2952            exp.SerdeProperties,
2953            **{  # type: ignore
2954                "expressions": self._parse_wrapped_properties(),
2955                "with": with_,
2956            },
2957        )
2958
2959    def _parse_row_format(
2960        self, match_row: bool = False
2961    ) -> t.Optional[exp.RowFormatSerdeProperty | exp.RowFormatDelimitedProperty]:
2962        if match_row and not self._match_pair(TokenType.ROW, TokenType.FORMAT):
2963            return None
2964
2965        if self._match_text_seq("SERDE"):
2966            this = self._parse_string()
2967
2968            serde_properties = self._parse_serde_properties()
2969
2970            return self.expression(
2971                exp.RowFormatSerdeProperty, this=this, serde_properties=serde_properties
2972            )
2973
2974        self._match_text_seq("DELIMITED")
2975
2976        kwargs = {}
2977
2978        if self._match_text_seq("FIELDS", "TERMINATED", "BY"):
2979            kwargs["fields"] = self._parse_string()
2980            if self._match_text_seq("ESCAPED", "BY"):
2981                kwargs["escaped"] = self._parse_string()
2982        if self._match_text_seq("COLLECTION", "ITEMS", "TERMINATED", "BY"):
2983            kwargs["collection_items"] = self._parse_string()
2984        if self._match_text_seq("MAP", "KEYS", "TERMINATED", "BY"):
2985            kwargs["map_keys"] = self._parse_string()
2986        if self._match_text_seq("LINES", "TERMINATED", "BY"):
2987            kwargs["lines"] = self._parse_string()
2988        if self._match_text_seq("NULL", "DEFINED", "AS"):
2989            kwargs["null"] = self._parse_string()
2990
2991        return self.expression(exp.RowFormatDelimitedProperty, **kwargs)  # type: ignore
2992
2993    def _parse_load(self) -> exp.LoadData | exp.Command:
2994        if self._match_text_seq("DATA"):
2995            local = self._match_text_seq("LOCAL")
2996            self._match_text_seq("INPATH")
2997            inpath = self._parse_string()
2998            overwrite = self._match(TokenType.OVERWRITE)
2999            self._match_pair(TokenType.INTO, TokenType.TABLE)
3000
3001            return self.expression(
3002                exp.LoadData,
3003                this=self._parse_table(schema=True),
3004                local=local,
3005                overwrite=overwrite,
3006                inpath=inpath,
3007                partition=self._parse_partition(),
3008                input_format=self._match_text_seq("INPUTFORMAT") and self._parse_string(),
3009                serde=self._match_text_seq("SERDE") and self._parse_string(),
3010            )
3011        return self._parse_as_command(self._prev)
3012
3013    def _parse_delete(self) -> exp.Delete:
3014        # This handles MySQL's "Multiple-Table Syntax"
3015        # https://dev.mysql.com/doc/refman/8.0/en/delete.html
3016        tables = None
3017        if not self._match(TokenType.FROM, advance=False):
3018            tables = self._parse_csv(self._parse_table) or None
3019
3020        returning = self._parse_returning()
3021
3022        return self.expression(
3023            exp.Delete,
3024            tables=tables,
3025            this=self._match(TokenType.FROM) and self._parse_table(joins=True),
3026            using=self._match(TokenType.USING) and self._parse_table(joins=True),
3027            cluster=self._match(TokenType.ON) and self._parse_on_property(),
3028            where=self._parse_where(),
3029            returning=returning or self._parse_returning(),
3030            limit=self._parse_limit(),
3031        )
3032
3033    def _parse_update(self) -> exp.Update:
3034        this = self._parse_table(joins=True, alias_tokens=self.UPDATE_ALIAS_TOKENS)
3035        expressions = self._match(TokenType.SET) and self._parse_csv(self._parse_equality)
3036        returning = self._parse_returning()
3037        return self.expression(
3038            exp.Update,
3039            **{  # type: ignore
3040                "this": this,
3041                "expressions": expressions,
3042                "from": self._parse_from(joins=True),
3043                "where": self._parse_where(),
3044                "returning": returning or self._parse_returning(),
3045                "order": self._parse_order(),
3046                "limit": self._parse_limit(),
3047            },
3048        )
3049
3050    def _parse_use(self) -> exp.Use:
3051        return self.expression(
3052            exp.Use,
3053            kind=self._parse_var_from_options(self.USABLES, raise_unmatched=False),
3054            this=self._parse_table(schema=False),
3055        )
3056
3057    def _parse_uncache(self) -> exp.Uncache:
3058        if not self._match(TokenType.TABLE):
3059            self.raise_error("Expecting TABLE after UNCACHE")
3060
3061        return self.expression(
3062            exp.Uncache, exists=self._parse_exists(), this=self._parse_table(schema=True)
3063        )
3064
3065    def _parse_cache(self) -> exp.Cache:
3066        lazy = self._match_text_seq("LAZY")
3067        self._match(TokenType.TABLE)
3068        table = self._parse_table(schema=True)
3069
3070        options = []
3071        if self._match_text_seq("OPTIONS"):
3072            self._match_l_paren()
3073            k = self._parse_string()
3074            self._match(TokenType.EQ)
3075            v = self._parse_string()
3076            options = [k, v]
3077            self._match_r_paren()
3078
3079        self._match(TokenType.ALIAS)
3080        return self.expression(
3081            exp.Cache,
3082            this=table,
3083            lazy=lazy,
3084            options=options,
3085            expression=self._parse_select(nested=True),
3086        )
3087
3088    def _parse_partition(self) -> t.Optional[exp.Partition]:
3089        if not self._match_texts(self.PARTITION_KEYWORDS):
3090            return None
3091
3092        return self.expression(
3093            exp.Partition,
3094            subpartition=self._prev.text.upper() == "SUBPARTITION",
3095            expressions=self._parse_wrapped_csv(self._parse_assignment),
3096        )
3097
3098    def _parse_value(self, values: bool = True) -> t.Optional[exp.Tuple]:
3099        def _parse_value_expression() -> t.Optional[exp.Expression]:
3100            if self.dialect.SUPPORTS_VALUES_DEFAULT and self._match(TokenType.DEFAULT):
3101                return exp.var(self._prev.text.upper())
3102            return self._parse_expression()
3103
3104        if self._match(TokenType.L_PAREN):
3105            expressions = self._parse_csv(_parse_value_expression)
3106            self._match_r_paren()
3107            return self.expression(exp.Tuple, expressions=expressions)
3108
3109        # In some dialects we can have VALUES 1, 2 which results in 1 column & 2 rows.
3110        expression = self._parse_expression()
3111        if expression:
3112            return self.expression(exp.Tuple, expressions=[expression])
3113        return None
3114
3115    def _parse_projections(self) -> t.List[exp.Expression]:
3116        return self._parse_expressions()
3117
3118    def _parse_wrapped_select(self, table: bool = False) -> t.Optional[exp.Expression]:
3119        if self._match_set((TokenType.PIVOT, TokenType.UNPIVOT)):
3120            this: t.Optional[exp.Expression] = self._parse_simplified_pivot(
3121                is_unpivot=self._prev.token_type == TokenType.UNPIVOT
3122            )
3123        elif self._match(TokenType.FROM):
3124            from_ = self._parse_from(skip_from_token=True)
3125            # Support parentheses for duckdb FROM-first syntax
3126            select = self._parse_select()
3127            if select:
3128                select.set("from", from_)
3129                this = select
3130            else:
3131                this = exp.select("*").from_(t.cast(exp.From, from_))
3132        else:
3133            this = (
3134                self._parse_table()
3135                if table
3136                else self._parse_select(nested=True, parse_set_operation=False)
3137            )
3138
3139            # Transform exp.Values into a exp.Table to pass through parse_query_modifiers
3140            # in case a modifier (e.g. join) is following
3141            if table and isinstance(this, exp.Values) and this.alias:
3142                alias = this.args["alias"].pop()
3143                this = exp.Table(this=this, alias=alias)
3144
3145            this = self._parse_query_modifiers(self._parse_set_operations(this))
3146
3147        return this
3148
3149    def _parse_select(
3150        self,
3151        nested: bool = False,
3152        table: bool = False,
3153        parse_subquery_alias: bool = True,
3154        parse_set_operation: bool = True,
3155    ) -> t.Optional[exp.Expression]:
3156        cte = self._parse_with()
3157
3158        if cte:
3159            this = self._parse_statement()
3160
3161            if not this:
3162                self.raise_error("Failed to parse any statement following CTE")
3163                return cte
3164
3165            if "with" in this.arg_types:
3166                this.set("with", cte)
3167            else:
3168                self.raise_error(f"{this.key} does not support CTE")
3169                this = cte
3170
3171            return this
3172
3173        # duckdb supports leading with FROM x
3174        from_ = self._parse_from() if self._match(TokenType.FROM, advance=False) else None
3175
3176        if self._match(TokenType.SELECT):
3177            comments = self._prev_comments
3178
3179            hint = self._parse_hint()
3180
3181            if self._next and not self._next.token_type == TokenType.DOT:
3182                all_ = self._match(TokenType.ALL)
3183                distinct = self._match_set(self.DISTINCT_TOKENS)
3184            else:
3185                all_, distinct = None, None
3186
3187            kind = (
3188                self._match(TokenType.ALIAS)
3189                and self._match_texts(("STRUCT", "VALUE"))
3190                and self._prev.text.upper()
3191            )
3192
3193            if distinct:
3194                distinct = self.expression(
3195                    exp.Distinct,
3196                    on=self._parse_value(values=False) if self._match(TokenType.ON) else None,
3197                )
3198
3199            if all_ and distinct:
3200                self.raise_error("Cannot specify both ALL and DISTINCT after SELECT")
3201
3202            operation_modifiers = []
3203            while self._curr and self._match_texts(self.OPERATION_MODIFIERS):
3204                operation_modifiers.append(exp.var(self._prev.text.upper()))
3205
3206            limit = self._parse_limit(top=True)
3207            projections = self._parse_projections()
3208
3209            this = self.expression(
3210                exp.Select,
3211                kind=kind,
3212                hint=hint,
3213                distinct=distinct,
3214                expressions=projections,
3215                limit=limit,
3216                operation_modifiers=operation_modifiers or None,
3217            )
3218            this.comments = comments
3219
3220            into = self._parse_into()
3221            if into:
3222                this.set("into", into)
3223
3224            if not from_:
3225                from_ = self._parse_from()
3226
3227            if from_:
3228                this.set("from", from_)
3229
3230            this = self._parse_query_modifiers(this)
3231        elif (table or nested) and self._match(TokenType.L_PAREN):
3232            this = self._parse_wrapped_select(table=table)
3233
3234            # We return early here so that the UNION isn't attached to the subquery by the
3235            # following call to _parse_set_operations, but instead becomes the parent node
3236            self._match_r_paren()
3237            return self._parse_subquery(this, parse_alias=parse_subquery_alias)
3238        elif self._match(TokenType.VALUES, advance=False):
3239            this = self._parse_derived_table_values()
3240        elif from_:
3241            this = exp.select("*").from_(from_.this, copy=False)
3242        elif self._match(TokenType.SUMMARIZE):
3243            table = self._match(TokenType.TABLE)
3244            this = self._parse_select() or self._parse_string() or self._parse_table()
3245            return self.expression(exp.Summarize, this=this, table=table)
3246        elif self._match(TokenType.DESCRIBE):
3247            this = self._parse_describe()
3248        elif self._match_text_seq("STREAM"):
3249            this = self._parse_function()
3250            if this:
3251                this = self.expression(exp.Stream, this=this)
3252            else:
3253                self._retreat(self._index - 1)
3254        else:
3255            this = None
3256
3257        return self._parse_set_operations(this) if parse_set_operation else this
3258
3259    def _parse_recursive_with_search(self) -> t.Optional[exp.RecursiveWithSearch]:
3260        self._match_text_seq("SEARCH")
3261
3262        kind = self._match_texts(self.RECURSIVE_CTE_SEARCH_KIND) and self._prev.text.upper()
3263
3264        if not kind:
3265            return None
3266
3267        self._match_text_seq("FIRST", "BY")
3268
3269        return self.expression(
3270            exp.RecursiveWithSearch,
3271            kind=kind,
3272            this=self._parse_id_var(),
3273            expression=self._match_text_seq("SET") and self._parse_id_var(),
3274            using=self._match_text_seq("USING") and self._parse_id_var(),
3275        )
3276
3277    def _parse_with(self, skip_with_token: bool = False) -> t.Optional[exp.With]:
3278        if not skip_with_token and not self._match(TokenType.WITH):
3279            return None
3280
3281        comments = self._prev_comments
3282        recursive = self._match(TokenType.RECURSIVE)
3283
3284        last_comments = None
3285        expressions = []
3286        while True:
3287            cte = self._parse_cte()
3288            if isinstance(cte, exp.CTE):
3289                expressions.append(cte)
3290                if last_comments:
3291                    cte.add_comments(last_comments)
3292
3293            if not self._match(TokenType.COMMA) and not self._match(TokenType.WITH):
3294                break
3295            else:
3296                self._match(TokenType.WITH)
3297
3298            last_comments = self._prev_comments
3299
3300        return self.expression(
3301            exp.With,
3302            comments=comments,
3303            expressions=expressions,
3304            recursive=recursive,
3305            search=self._parse_recursive_with_search(),
3306        )
3307
3308    def _parse_cte(self) -> t.Optional[exp.CTE]:
3309        index = self._index
3310
3311        alias = self._parse_table_alias(self.ID_VAR_TOKENS)
3312        if not alias or not alias.this:
3313            self.raise_error("Expected CTE to have alias")
3314
3315        if not self._match(TokenType.ALIAS) and not self.OPTIONAL_ALIAS_TOKEN_CTE:
3316            self._retreat(index)
3317            return None
3318
3319        comments = self._prev_comments
3320
3321        if self._match_text_seq("NOT", "MATERIALIZED"):
3322            materialized = False
3323        elif self._match_text_seq("MATERIALIZED"):
3324            materialized = True
3325        else:
3326            materialized = None
3327
3328        cte = self.expression(
3329            exp.CTE,
3330            this=self._parse_wrapped(self._parse_statement),
3331            alias=alias,
3332            materialized=materialized,
3333            comments=comments,
3334        )
3335
3336        if isinstance(cte.this, exp.Values):
3337            cte.set("this", exp.select("*").from_(exp.alias_(cte.this, "_values", table=True)))
3338
3339        return cte
3340
3341    def _parse_table_alias(
3342        self, alias_tokens: t.Optional[t.Collection[TokenType]] = None
3343    ) -> t.Optional[exp.TableAlias]:
3344        # In some dialects, LIMIT and OFFSET can act as both identifiers and keywords (clauses)
3345        # so this section tries to parse the clause version and if it fails, it treats the token
3346        # as an identifier (alias)
3347        if self._can_parse_limit_or_offset():
3348            return None
3349
3350        any_token = self._match(TokenType.ALIAS)
3351        alias = (
3352            self._parse_id_var(any_token=any_token, tokens=alias_tokens or self.TABLE_ALIAS_TOKENS)
3353            or self._parse_string_as_identifier()
3354        )
3355
3356        index = self._index
3357        if self._match(TokenType.L_PAREN):
3358            columns = self._parse_csv(self._parse_function_parameter)
3359            self._match_r_paren() if columns else self._retreat(index)
3360        else:
3361            columns = None
3362
3363        if not alias and not columns:
3364            return None
3365
3366        table_alias = self.expression(exp.TableAlias, this=alias, columns=columns)
3367
3368        # We bubble up comments from the Identifier to the TableAlias
3369        if isinstance(alias, exp.Identifier):
3370            table_alias.add_comments(alias.pop_comments())
3371
3372        return table_alias
3373
3374    def _parse_subquery(
3375        self, this: t.Optional[exp.Expression], parse_alias: bool = True
3376    ) -> t.Optional[exp.Subquery]:
3377        if not this:
3378            return None
3379
3380        return self.expression(
3381            exp.Subquery,
3382            this=this,
3383            pivots=self._parse_pivots(),
3384            alias=self._parse_table_alias() if parse_alias else None,
3385            sample=self._parse_table_sample(),
3386        )
3387
3388    def _implicit_unnests_to_explicit(self, this: E) -> E:
3389        from sqlglot.optimizer.normalize_identifiers import normalize_identifiers as _norm
3390
3391        refs = {_norm(this.args["from"].this.copy(), dialect=self.dialect).alias_or_name}
3392        for i, join in enumerate(this.args.get("joins") or []):
3393            table = join.this
3394            normalized_table = table.copy()
3395            normalized_table.meta["maybe_column"] = True
3396            normalized_table = _norm(normalized_table, dialect=self.dialect)
3397
3398            if isinstance(table, exp.Table) and not join.args.get("on"):
3399                if normalized_table.parts[0].name in refs:
3400                    table_as_column = table.to_column()
3401                    unnest = exp.Unnest(expressions=[table_as_column])
3402
3403                    # Table.to_column creates a parent Alias node that we want to convert to
3404                    # a TableAlias and attach to the Unnest, so it matches the parser's output
3405                    if isinstance(table.args.get("alias"), exp.TableAlias):
3406                        table_as_column.replace(table_as_column.this)
3407                        exp.alias_(unnest, None, table=[table.args["alias"].this], copy=False)
3408
3409                    table.replace(unnest)
3410
3411            refs.add(normalized_table.alias_or_name)
3412
3413        return this
3414
3415    def _parse_query_modifiers(
3416        self, this: t.Optional[exp.Expression]
3417    ) -> t.Optional[exp.Expression]:
3418        if isinstance(this, self.MODIFIABLES):
3419            for join in self._parse_joins():
3420                this.append("joins", join)
3421            for lateral in iter(self._parse_lateral, None):
3422                this.append("laterals", lateral)
3423
3424            while True:
3425                if self._match_set(self.QUERY_MODIFIER_PARSERS, advance=False):
3426                    parser = self.QUERY_MODIFIER_PARSERS[self._curr.token_type]
3427                    key, expression = parser(self)
3428
3429                    if expression:
3430                        this.set(key, expression)
3431                        if key == "limit":
3432                            offset = expression.args.pop("offset", None)
3433
3434                            if offset:
3435                                offset = exp.Offset(expression=offset)
3436                                this.set("offset", offset)
3437
3438                                limit_by_expressions = expression.expressions
3439                                expression.set("expressions", None)
3440                                offset.set("expressions", limit_by_expressions)
3441                        continue
3442                break
3443
3444        if self.SUPPORTS_IMPLICIT_UNNEST and this and this.args.get("from"):
3445            this = self._implicit_unnests_to_explicit(this)
3446
3447        return this
3448
3449    def _parse_hint_fallback_to_string(self) -> t.Optional[exp.Hint]:
3450        start = self._curr
3451        while self._curr:
3452            self._advance()
3453
3454        end = self._tokens[self._index - 1]
3455        return exp.Hint(expressions=[self._find_sql(start, end)])
3456
3457    def _parse_hint_function_call(self) -> t.Optional[exp.Expression]:
3458        return self._parse_function_call()
3459
3460    def _parse_hint_body(self) -> t.Optional[exp.Hint]:
3461        start_index = self._index
3462        should_fallback_to_string = False
3463
3464        hints = []
3465        try:
3466            for hint in iter(
3467                lambda: self._parse_csv(
3468                    lambda: self._parse_hint_function_call() or self._parse_var(upper=True),
3469                ),
3470                [],
3471            ):
3472                hints.extend(hint)
3473        except ParseError:
3474            should_fallback_to_string = True
3475
3476        if should_fallback_to_string or self._curr:
3477            self._retreat(start_index)
3478            return self._parse_hint_fallback_to_string()
3479
3480        return self.expression(exp.Hint, expressions=hints)
3481
3482    def _parse_hint(self) -> t.Optional[exp.Hint]:
3483        if self._match(TokenType.HINT) and self._prev_comments:
3484            return exp.maybe_parse(self._prev_comments[0], into=exp.Hint, dialect=self.dialect)
3485
3486        return None
3487
3488    def _parse_into(self) -> t.Optional[exp.Into]:
3489        if not self._match(TokenType.INTO):
3490            return None
3491
3492        temp = self._match(TokenType.TEMPORARY)
3493        unlogged = self._match_text_seq("UNLOGGED")
3494        self._match(TokenType.TABLE)
3495
3496        return self.expression(
3497            exp.Into, this=self._parse_table(schema=True), temporary=temp, unlogged=unlogged
3498        )
3499
3500    def _parse_from(
3501        self, joins: bool = False, skip_from_token: bool = False
3502    ) -> t.Optional[exp.From]:
3503        if not skip_from_token and not self._match(TokenType.FROM):
3504            return None
3505
3506        return self.expression(
3507            exp.From, comments=self._prev_comments, this=self._parse_table(joins=joins)
3508        )
3509
3510    def _parse_match_recognize_measure(self) -> exp.MatchRecognizeMeasure:
3511        return self.expression(
3512            exp.MatchRecognizeMeasure,
3513            window_frame=self._match_texts(("FINAL", "RUNNING")) and self._prev.text.upper(),
3514            this=self._parse_expression(),
3515        )
3516
3517    def _parse_match_recognize(self) -> t.Optional[exp.MatchRecognize]:
3518        if not self._match(TokenType.MATCH_RECOGNIZE):
3519            return None
3520
3521        self._match_l_paren()
3522
3523        partition = self._parse_partition_by()
3524        order = self._parse_order()
3525
3526        measures = (
3527            self._parse_csv(self._parse_match_recognize_measure)
3528            if self._match_text_seq("MEASURES")
3529            else None
3530        )
3531
3532        if self._match_text_seq("ONE", "ROW", "PER", "MATCH"):
3533            rows = exp.var("ONE ROW PER MATCH")
3534        elif self._match_text_seq("ALL", "ROWS", "PER", "MATCH"):
3535            text = "ALL ROWS PER MATCH"
3536            if self._match_text_seq("SHOW", "EMPTY", "MATCHES"):
3537                text += " SHOW EMPTY MATCHES"
3538            elif self._match_text_seq("OMIT", "EMPTY", "MATCHES"):
3539                text += " OMIT EMPTY MATCHES"
3540            elif self._match_text_seq("WITH", "UNMATCHED", "ROWS"):
3541                text += " WITH UNMATCHED ROWS"
3542            rows = exp.var(text)
3543        else:
3544            rows = None
3545
3546        if self._match_text_seq("AFTER", "MATCH", "SKIP"):
3547            text = "AFTER MATCH SKIP"
3548            if self._match_text_seq("PAST", "LAST", "ROW"):
3549                text += " PAST LAST ROW"
3550            elif self._match_text_seq("TO", "NEXT", "ROW"):
3551                text += " TO NEXT ROW"
3552            elif self._match_text_seq("TO", "FIRST"):
3553                text += f" TO FIRST {self._advance_any().text}"  # type: ignore
3554            elif self._match_text_seq("TO", "LAST"):
3555                text += f" TO LAST {self._advance_any().text}"  # type: ignore
3556            after = exp.var(text)
3557        else:
3558            after = None
3559
3560        if self._match_text_seq("PATTERN"):
3561            self._match_l_paren()
3562
3563            if not self._curr:
3564                self.raise_error("Expecting )", self._curr)
3565
3566            paren = 1
3567            start = self._curr
3568
3569            while self._curr and paren > 0:
3570                if self._curr.token_type == TokenType.L_PAREN:
3571                    paren += 1
3572                if self._curr.token_type == TokenType.R_PAREN:
3573                    paren -= 1
3574
3575                end = self._prev
3576                self._advance()
3577
3578            if paren > 0:
3579                self.raise_error("Expecting )", self._curr)
3580
3581            pattern = exp.var(self._find_sql(start, end))
3582        else:
3583            pattern = None
3584
3585        define = (
3586            self._parse_csv(self._parse_name_as_expression)
3587            if self._match_text_seq("DEFINE")
3588            else None
3589        )
3590
3591        self._match_r_paren()
3592
3593        return self.expression(
3594            exp.MatchRecognize,
3595            partition_by=partition,
3596            order=order,
3597            measures=measures,
3598            rows=rows,
3599            after=after,
3600            pattern=pattern,
3601            define=define,
3602            alias=self._parse_table_alias(),
3603        )
3604
3605    def _parse_lateral(self) -> t.Optional[exp.Lateral]:
3606        cross_apply = self._match_pair(TokenType.CROSS, TokenType.APPLY)
3607        if not cross_apply and self._match_pair(TokenType.OUTER, TokenType.APPLY):
3608            cross_apply = False
3609
3610        if cross_apply is not None:
3611            this = self._parse_select(table=True)
3612            view = None
3613            outer = None
3614        elif self._match(TokenType.LATERAL):
3615            this = self._parse_select(table=True)
3616            view = self._match(TokenType.VIEW)
3617            outer = self._match(TokenType.OUTER)
3618        else:
3619            return None
3620
3621        if not this:
3622            this = (
3623                self._parse_unnest()
3624                or self._parse_function()
3625                or self._parse_id_var(any_token=False)
3626            )
3627
3628            while self._match(TokenType.DOT):
3629                this = exp.Dot(
3630                    this=this,
3631                    expression=self._parse_function() or self._parse_id_var(any_token=False),
3632                )
3633
3634        ordinality: t.Optional[bool] = None
3635
3636        if view:
3637            table = self._parse_id_var(any_token=False)
3638            columns = self._parse_csv(self._parse_id_var) if self._match(TokenType.ALIAS) else []
3639            table_alias: t.Optional[exp.TableAlias] = self.expression(
3640                exp.TableAlias, this=table, columns=columns
3641            )
3642        elif isinstance(this, (exp.Subquery, exp.Unnest)) and this.alias:
3643            # We move the alias from the lateral's child node to the lateral itself
3644            table_alias = this.args["alias"].pop()
3645        else:
3646            ordinality = self._match_pair(TokenType.WITH, TokenType.ORDINALITY)
3647            table_alias = self._parse_table_alias()
3648
3649        return self.expression(
3650            exp.Lateral,
3651            this=this,
3652            view=view,
3653            outer=outer,
3654            alias=table_alias,
3655            cross_apply=cross_apply,
3656            ordinality=ordinality,
3657        )
3658
3659    def _parse_join_parts(
3660        self,
3661    ) -> t.Tuple[t.Optional[Token], t.Optional[Token], t.Optional[Token]]:
3662        return (
3663            self._match_set(self.JOIN_METHODS) and self._prev,
3664            self._match_set(self.JOIN_SIDES) and self._prev,
3665            self._match_set(self.JOIN_KINDS) and self._prev,
3666        )
3667
3668    def _parse_using_identifiers(self) -> t.List[exp.Expression]:
3669        def _parse_column_as_identifier() -> t.Optional[exp.Expression]:
3670            this = self._parse_column()
3671            if isinstance(this, exp.Column):
3672                return this.this
3673            return this
3674
3675        return self._parse_wrapped_csv(_parse_column_as_identifier, optional=True)
3676
3677    def _parse_join(
3678        self, skip_join_token: bool = False, parse_bracket: bool = False
3679    ) -> t.Optional[exp.Join]:
3680        if self._match(TokenType.COMMA):
3681            table = self._try_parse(self._parse_table)
3682            if table:
3683                return self.expression(exp.Join, this=table)
3684            return None
3685
3686        index = self._index
3687        method, side, kind = self._parse_join_parts()
3688        hint = self._prev.text if self._match_texts(self.JOIN_HINTS) else None
3689        join = self._match(TokenType.JOIN) or (kind and kind.token_type == TokenType.STRAIGHT_JOIN)
3690
3691        if not skip_join_token and not join:
3692            self._retreat(index)
3693            kind = None
3694            method = None
3695            side = None
3696
3697        outer_apply = self._match_pair(TokenType.OUTER, TokenType.APPLY, False)
3698        cross_apply = self._match_pair(TokenType.CROSS, TokenType.APPLY, False)
3699
3700        if not skip_join_token and not join and not outer_apply and not cross_apply:
3701            return None
3702
3703        kwargs: t.Dict[str, t.Any] = {"this": self._parse_table(parse_bracket=parse_bracket)}
3704        if kind and kind.token_type == TokenType.ARRAY and self._match(TokenType.COMMA):
3705            kwargs["expressions"] = self._parse_csv(
3706                lambda: self._parse_table(parse_bracket=parse_bracket)
3707            )
3708
3709        if method:
3710            kwargs["method"] = method.text
3711        if side:
3712            kwargs["side"] = side.text
3713        if kind:
3714            kwargs["kind"] = kind.text
3715        if hint:
3716            kwargs["hint"] = hint
3717
3718        if self._match(TokenType.MATCH_CONDITION):
3719            kwargs["match_condition"] = self._parse_wrapped(self._parse_comparison)
3720
3721        if self._match(TokenType.ON):
3722            kwargs["on"] = self._parse_assignment()
3723        elif self._match(TokenType.USING):
3724            kwargs["using"] = self._parse_using_identifiers()
3725        elif (
3726            not (outer_apply or cross_apply)
3727            and not isinstance(kwargs["this"], exp.Unnest)
3728            and not (kind and kind.token_type in (TokenType.CROSS, TokenType.ARRAY))
3729        ):
3730            index = self._index
3731            joins: t.Optional[list] = list(self._parse_joins())
3732
3733            if joins and self._match(TokenType.ON):
3734                kwargs["on"] = self._parse_assignment()
3735            elif joins and self._match(TokenType.USING):
3736                kwargs["using"] = self._parse_using_identifiers()
3737            else:
3738                joins = None
3739                self._retreat(index)
3740
3741            kwargs["this"].set("joins", joins if joins else None)
3742
3743        kwargs["pivots"] = self._parse_pivots()
3744
3745        comments = [c for token in (method, side, kind) if token for c in token.comments]
3746        return self.expression(exp.Join, comments=comments, **kwargs)
3747
3748    def _parse_opclass(self) -> t.Optional[exp.Expression]:
3749        this = self._parse_assignment()
3750
3751        if self._match_texts(self.OPCLASS_FOLLOW_KEYWORDS, advance=False):
3752            return this
3753
3754        if not self._match_set(self.OPTYPE_FOLLOW_TOKENS, advance=False):
3755            return self.expression(exp.Opclass, this=this, expression=self._parse_table_parts())
3756
3757        return this
3758
3759    def _parse_index_params(self) -> exp.IndexParameters:
3760        using = self._parse_var(any_token=True) if self._match(TokenType.USING) else None
3761
3762        if self._match(TokenType.L_PAREN, advance=False):
3763            columns = self._parse_wrapped_csv(self._parse_with_operator)
3764        else:
3765            columns = None
3766
3767        include = self._parse_wrapped_id_vars() if self._match_text_seq("INCLUDE") else None
3768        partition_by = self._parse_partition_by()
3769        with_storage = self._match(TokenType.WITH) and self._parse_wrapped_properties()
3770        tablespace = (
3771            self._parse_var(any_token=True)
3772            if self._match_text_seq("USING", "INDEX", "TABLESPACE")
3773            else None
3774        )
3775        where = self._parse_where()
3776
3777        on = self._parse_field() if self._match(TokenType.ON) else None
3778
3779        return self.expression(
3780            exp.IndexParameters,
3781            using=using,
3782            columns=columns,
3783            include=include,
3784            partition_by=partition_by,
3785            where=where,
3786            with_storage=with_storage,
3787            tablespace=tablespace,
3788            on=on,
3789        )
3790
3791    def _parse_index(
3792        self, index: t.Optional[exp.Expression] = None, anonymous: bool = False
3793    ) -> t.Optional[exp.Index]:
3794        if index or anonymous:
3795            unique = None
3796            primary = None
3797            amp = None
3798
3799            self._match(TokenType.ON)
3800            self._match(TokenType.TABLE)  # hive
3801            table = self._parse_table_parts(schema=True)
3802        else:
3803            unique = self._match(TokenType.UNIQUE)
3804            primary = self._match_text_seq("PRIMARY")
3805            amp = self._match_text_seq("AMP")
3806
3807            if not self._match(TokenType.INDEX):
3808                return None
3809
3810            index = self._parse_id_var()
3811            table = None
3812
3813        params = self._parse_index_params()
3814
3815        return self.expression(
3816            exp.Index,
3817            this=index,
3818            table=table,
3819            unique=unique,
3820            primary=primary,
3821            amp=amp,
3822            params=params,
3823        )
3824
3825    def _parse_table_hints(self) -> t.Optional[t.List[exp.Expression]]:
3826        hints: t.List[exp.Expression] = []
3827        if self._match_pair(TokenType.WITH, TokenType.L_PAREN):
3828            # https://learn.microsoft.com/en-us/sql/t-sql/queries/hints-transact-sql-table?view=sql-server-ver16
3829            hints.append(
3830                self.expression(
3831                    exp.WithTableHint,
3832                    expressions=self._parse_csv(
3833                        lambda: self._parse_function() or self._parse_var(any_token=True)
3834                    ),
3835                )
3836            )
3837            self._match_r_paren()
3838        else:
3839            # https://dev.mysql.com/doc/refman/8.0/en/index-hints.html
3840            while self._match_set(self.TABLE_INDEX_HINT_TOKENS):
3841                hint = exp.IndexTableHint(this=self._prev.text.upper())
3842
3843                self._match_set((TokenType.INDEX, TokenType.KEY))
3844                if self._match(TokenType.FOR):
3845                    hint.set("target", self._advance_any() and self._prev.text.upper())
3846
3847                hint.set("expressions", self._parse_wrapped_id_vars())
3848                hints.append(hint)
3849
3850        return hints or None
3851
3852    def _parse_table_part(self, schema: bool = False) -> t.Optional[exp.Expression]:
3853        return (
3854            (not schema and self._parse_function(optional_parens=False))
3855            or self._parse_id_var(any_token=False)
3856            or self._parse_string_as_identifier()
3857            or self._parse_placeholder()
3858        )
3859
3860    def _parse_table_parts(
3861        self, schema: bool = False, is_db_reference: bool = False, wildcard: bool = False
3862    ) -> exp.Table:
3863        catalog = None
3864        db = None
3865        table: t.Optional[exp.Expression | str] = self._parse_table_part(schema=schema)
3866
3867        while self._match(TokenType.DOT):
3868            if catalog:
3869                # This allows nesting the table in arbitrarily many dot expressions if needed
3870                table = self.expression(
3871                    exp.Dot, this=table, expression=self._parse_table_part(schema=schema)
3872                )
3873            else:
3874                catalog = db
3875                db = table
3876                # "" used for tsql FROM a..b case
3877                table = self._parse_table_part(schema=schema) or ""
3878
3879        if (
3880            wildcard
3881            and self._is_connected()
3882            and (isinstance(table, exp.Identifier) or not table)
3883            and self._match(TokenType.STAR)
3884        ):
3885            if isinstance(table, exp.Identifier):
3886                table.args["this"] += "*"
3887            else:
3888                table = exp.Identifier(this="*")
3889
3890        # We bubble up comments from the Identifier to the Table
3891        comments = table.pop_comments() if isinstance(table, exp.Expression) else None
3892
3893        if is_db_reference:
3894            catalog = db
3895            db = table
3896            table = None
3897
3898        if not table and not is_db_reference:
3899            self.raise_error(f"Expected table name but got {self._curr}")
3900        if not db and is_db_reference:
3901            self.raise_error(f"Expected database name but got {self._curr}")
3902
3903        table = self.expression(
3904            exp.Table,
3905            comments=comments,
3906            this=table,
3907            db=db,
3908            catalog=catalog,
3909        )
3910
3911        changes = self._parse_changes()
3912        if changes:
3913            table.set("changes", changes)
3914
3915        at_before = self._parse_historical_data()
3916        if at_before:
3917            table.set("when", at_before)
3918
3919        pivots = self._parse_pivots()
3920        if pivots:
3921            table.set("pivots", pivots)
3922
3923        return table
3924
3925    def _parse_table(
3926        self,
3927        schema: bool = False,
3928        joins: bool = False,
3929        alias_tokens: t.Optional[t.Collection[TokenType]] = None,
3930        parse_bracket: bool = False,
3931        is_db_reference: bool = False,
3932        parse_partition: bool = False,
3933    ) -> t.Optional[exp.Expression]:
3934        lateral = self._parse_lateral()
3935        if lateral:
3936            return lateral
3937
3938        unnest = self._parse_unnest()
3939        if unnest:
3940            return unnest
3941
3942        values = self._parse_derived_table_values()
3943        if values:
3944            return values
3945
3946        subquery = self._parse_select(table=True)
3947        if subquery:
3948            if not subquery.args.get("pivots"):
3949                subquery.set("pivots", self._parse_pivots())
3950            return subquery
3951
3952        bracket = parse_bracket and self._parse_bracket(None)
3953        bracket = self.expression(exp.Table, this=bracket) if bracket else None
3954
3955        rows_from = self._match_text_seq("ROWS", "FROM") and self._parse_wrapped_csv(
3956            self._parse_table
3957        )
3958        rows_from = self.expression(exp.Table, rows_from=rows_from) if rows_from else None
3959
3960        only = self._match(TokenType.ONLY)
3961
3962        this = t.cast(
3963            exp.Expression,
3964            bracket
3965            or rows_from
3966            or self._parse_bracket(
3967                self._parse_table_parts(schema=schema, is_db_reference=is_db_reference)
3968            ),
3969        )
3970
3971        if only:
3972            this.set("only", only)
3973
3974        # Postgres supports a wildcard (table) suffix operator, which is a no-op in this context
3975        self._match_text_seq("*")
3976
3977        parse_partition = parse_partition or self.SUPPORTS_PARTITION_SELECTION
3978        if parse_partition and self._match(TokenType.PARTITION, advance=False):
3979            this.set("partition", self._parse_partition())
3980
3981        if schema:
3982            return self._parse_schema(this=this)
3983
3984        version = self._parse_version()
3985
3986        if version:
3987            this.set("version", version)
3988
3989        if self.dialect.ALIAS_POST_TABLESAMPLE:
3990            this.set("sample", self._parse_table_sample())
3991
3992        alias = self._parse_table_alias(alias_tokens=alias_tokens or self.TABLE_ALIAS_TOKENS)
3993        if alias:
3994            this.set("alias", alias)
3995
3996        if isinstance(this, exp.Table) and self._match_text_seq("AT"):
3997            return self.expression(
3998                exp.AtIndex, this=this.to_column(copy=False), expression=self._parse_id_var()
3999            )
4000
4001        this.set("hints", self._parse_table_hints())
4002
4003        if not this.args.get("pivots"):
4004            this.set("pivots", self._parse_pivots())
4005
4006        if not self.dialect.ALIAS_POST_TABLESAMPLE:
4007            this.set("sample", self._parse_table_sample())
4008
4009        if joins:
4010            for join in self._parse_joins():
4011                this.append("joins", join)
4012
4013        if self._match_pair(TokenType.WITH, TokenType.ORDINALITY):
4014            this.set("ordinality", True)
4015            this.set("alias", self._parse_table_alias())
4016
4017        return this
4018
4019    def _parse_version(self) -> t.Optional[exp.Version]:
4020        if self._match(TokenType.TIMESTAMP_SNAPSHOT):
4021            this = "TIMESTAMP"
4022        elif self._match(TokenType.VERSION_SNAPSHOT):
4023            this = "VERSION"
4024        else:
4025            return None
4026
4027        if self._match_set((TokenType.FROM, TokenType.BETWEEN)):
4028            kind = self._prev.text.upper()
4029            start = self._parse_bitwise()
4030            self._match_texts(("TO", "AND"))
4031            end = self._parse_bitwise()
4032            expression: t.Optional[exp.Expression] = self.expression(
4033                exp.Tuple, expressions=[start, end]
4034            )
4035        elif self._match_text_seq("CONTAINED", "IN"):
4036            kind = "CONTAINED IN"
4037            expression = self.expression(
4038                exp.Tuple, expressions=self._parse_wrapped_csv(self._parse_bitwise)
4039            )
4040        elif self._match(TokenType.ALL):
4041            kind = "ALL"
4042            expression = None
4043        else:
4044            self._match_text_seq("AS", "OF")
4045            kind = "AS OF"
4046            expression = self._parse_type()
4047
4048        return self.expression(exp.Version, this=this, expression=expression, kind=kind)
4049
4050    def _parse_historical_data(self) -> t.Optional[exp.HistoricalData]:
4051        # https://docs.snowflake.com/en/sql-reference/constructs/at-before
4052        index = self._index
4053        historical_data = None
4054        if self._match_texts(self.HISTORICAL_DATA_PREFIX):
4055            this = self._prev.text.upper()
4056            kind = (
4057                self._match(TokenType.L_PAREN)
4058                and self._match_texts(self.HISTORICAL_DATA_KIND)
4059                and self._prev.text.upper()
4060            )
4061            expression = self._match(TokenType.FARROW) and self._parse_bitwise()
4062
4063            if expression:
4064                self._match_r_paren()
4065                historical_data = self.expression(
4066                    exp.HistoricalData, this=this, kind=kind, expression=expression
4067                )
4068            else:
4069                self._retreat(index)
4070
4071        return historical_data
4072
4073    def _parse_changes(self) -> t.Optional[exp.Changes]:
4074        if not self._match_text_seq("CHANGES", "(", "INFORMATION", "=>"):
4075            return None
4076
4077        information = self._parse_var(any_token=True)
4078        self._match_r_paren()
4079
4080        return self.expression(
4081            exp.Changes,
4082            information=information,
4083            at_before=self._parse_historical_data(),
4084            end=self._parse_historical_data(),
4085        )
4086
4087    def _parse_unnest(self, with_alias: bool = True) -> t.Optional[exp.Unnest]:
4088        if not self._match(TokenType.UNNEST):
4089            return None
4090
4091        expressions = self._parse_wrapped_csv(self._parse_equality)
4092        offset = self._match_pair(TokenType.WITH, TokenType.ORDINALITY)
4093
4094        alias = self._parse_table_alias() if with_alias else None
4095
4096        if alias:
4097            if self.dialect.UNNEST_COLUMN_ONLY:
4098                if alias.args.get("columns"):
4099                    self.raise_error("Unexpected extra column alias in unnest.")
4100
4101                alias.set("columns", [alias.this])
4102                alias.set("this", None)
4103
4104            columns = alias.args.get("columns") or []
4105            if offset and len(expressions) < len(columns):
4106                offset = columns.pop()
4107
4108        if not offset and self._match_pair(TokenType.WITH, TokenType.OFFSET):
4109            self._match(TokenType.ALIAS)
4110            offset = self._parse_id_var(
4111                any_token=False, tokens=self.UNNEST_OFFSET_ALIAS_TOKENS
4112            ) or exp.to_identifier("offset")
4113
4114        return self.expression(exp.Unnest, expressions=expressions, alias=alias, offset=offset)
4115
4116    def _parse_derived_table_values(self) -> t.Optional[exp.Values]:
4117        is_derived = self._match_pair(TokenType.L_PAREN, TokenType.VALUES)
4118        if not is_derived and not (
4119            # ClickHouse's `FORMAT Values` is equivalent to `VALUES`
4120            self._match_text_seq("VALUES") or self._match_text_seq("FORMAT", "VALUES")
4121        ):
4122            return None
4123
4124        expressions = self._parse_csv(self._parse_value)
4125        alias = self._parse_table_alias()
4126
4127        if is_derived:
4128            self._match_r_paren()
4129
4130        return self.expression(
4131            exp.Values, expressions=expressions, alias=alias or self._parse_table_alias()
4132        )
4133
4134    def _parse_table_sample(self, as_modifier: bool = False) -> t.Optional[exp.TableSample]:
4135        if not self._match(TokenType.TABLE_SAMPLE) and not (
4136            as_modifier and self._match_text_seq("USING", "SAMPLE")
4137        ):
4138            return None
4139
4140        bucket_numerator = None
4141        bucket_denominator = None
4142        bucket_field = None
4143        percent = None
4144        size = None
4145        seed = None
4146
4147        method = self._parse_var(tokens=(TokenType.ROW,), upper=True)
4148        matched_l_paren = self._match(TokenType.L_PAREN)
4149
4150        if self.TABLESAMPLE_CSV:
4151            num = None
4152            expressions = self._parse_csv(self._parse_primary)
4153        else:
4154            expressions = None
4155            num = (
4156                self._parse_factor()
4157                if self._match(TokenType.NUMBER, advance=False)
4158                else self._parse_primary() or self._parse_placeholder()
4159            )
4160
4161        if self._match_text_seq("BUCKET"):
4162            bucket_numerator = self._parse_number()
4163            self._match_text_seq("OUT", "OF")
4164            bucket_denominator = bucket_denominator = self._parse_number()
4165            self._match(TokenType.ON)
4166            bucket_field = self._parse_field()
4167        elif self._match_set((TokenType.PERCENT, TokenType.MOD)):
4168            percent = num
4169        elif self._match(TokenType.ROWS) or not self.dialect.TABLESAMPLE_SIZE_IS_PERCENT:
4170            size = num
4171        else:
4172            percent = num
4173
4174        if matched_l_paren:
4175            self._match_r_paren()
4176
4177        if self._match(TokenType.L_PAREN):
4178            method = self._parse_var(upper=True)
4179            seed = self._match(TokenType.COMMA) and self._parse_number()
4180            self._match_r_paren()
4181        elif self._match_texts(("SEED", "REPEATABLE")):
4182            seed = self._parse_wrapped(self._parse_number)
4183
4184        if not method and self.DEFAULT_SAMPLING_METHOD:
4185            method = exp.var(self.DEFAULT_SAMPLING_METHOD)
4186
4187        return self.expression(
4188            exp.TableSample,
4189            expressions=expressions,
4190            method=method,
4191            bucket_numerator=bucket_numerator,
4192            bucket_denominator=bucket_denominator,
4193            bucket_field=bucket_field,
4194            percent=percent,
4195            size=size,
4196            seed=seed,
4197        )
4198
4199    def _parse_pivots(self) -> t.Optional[t.List[exp.Pivot]]:
4200        return list(iter(self._parse_pivot, None)) or None
4201
4202    def _parse_joins(self) -> t.Iterator[exp.Join]:
4203        return iter(self._parse_join, None)
4204
4205    def _parse_unpivot_columns(self) -> t.Optional[exp.UnpivotColumns]:
4206        if not self._match(TokenType.INTO):
4207            return None
4208
4209        return self.expression(
4210            exp.UnpivotColumns,
4211            this=self._match_text_seq("NAME") and self._parse_column(),
4212            expressions=self._match_text_seq("VALUE") and self._parse_csv(self._parse_column),
4213        )
4214
4215    # https://duckdb.org/docs/sql/statements/pivot
4216    def _parse_simplified_pivot(self, is_unpivot: t.Optional[bool] = None) -> exp.Pivot:
4217        def _parse_on() -> t.Optional[exp.Expression]:
4218            this = self._parse_bitwise()
4219
4220            if self._match(TokenType.IN):
4221                # PIVOT ... ON col IN (row_val1, row_val2)
4222                return self._parse_in(this)
4223            if self._match(TokenType.ALIAS, advance=False):
4224                # UNPIVOT ... ON (col1, col2, col3) AS row_val
4225                return self._parse_alias(this)
4226
4227            return this
4228
4229        this = self._parse_table()
4230        expressions = self._match(TokenType.ON) and self._parse_csv(_parse_on)
4231        into = self._parse_unpivot_columns()
4232        using = self._match(TokenType.USING) and self._parse_csv(
4233            lambda: self._parse_alias(self._parse_function())
4234        )
4235        group = self._parse_group()
4236
4237        return self.expression(
4238            exp.Pivot,
4239            this=this,
4240            expressions=expressions,
4241            using=using,
4242            group=group,
4243            unpivot=is_unpivot,
4244            into=into,
4245        )
4246
4247    def _parse_pivot_in(self) -> exp.In:
4248        def _parse_aliased_expression() -> t.Optional[exp.Expression]:
4249            this = self._parse_select_or_expression()
4250
4251            self._match(TokenType.ALIAS)
4252            alias = self._parse_bitwise()
4253            if alias:
4254                if isinstance(alias, exp.Column) and not alias.db:
4255                    alias = alias.this
4256                return self.expression(exp.PivotAlias, this=this, alias=alias)
4257
4258            return this
4259
4260        value = self._parse_column()
4261
4262        if not self._match_pair(TokenType.IN, TokenType.L_PAREN):
4263            self.raise_error("Expecting IN (")
4264
4265        if self._match(TokenType.ANY):
4266            exprs: t.List[exp.Expression] = ensure_list(exp.PivotAny(this=self._parse_order()))
4267        else:
4268            exprs = self._parse_csv(_parse_aliased_expression)
4269
4270        self._match_r_paren()
4271        return self.expression(exp.In, this=value, expressions=exprs)
4272
4273    def _parse_pivot(self) -> t.Optional[exp.Pivot]:
4274        index = self._index
4275        include_nulls = None
4276
4277        if self._match(TokenType.PIVOT):
4278            unpivot = False
4279        elif self._match(TokenType.UNPIVOT):
4280            unpivot = True
4281
4282            # https://docs.databricks.com/en/sql/language-manual/sql-ref-syntax-qry-select-unpivot.html#syntax
4283            if self._match_text_seq("INCLUDE", "NULLS"):
4284                include_nulls = True
4285            elif self._match_text_seq("EXCLUDE", "NULLS"):
4286                include_nulls = False
4287        else:
4288            return None
4289
4290        expressions = []
4291
4292        if not self._match(TokenType.L_PAREN):
4293            self._retreat(index)
4294            return None
4295
4296        if unpivot:
4297            expressions = self._parse_csv(self._parse_column)
4298        else:
4299            expressions = self._parse_csv(lambda: self._parse_alias(self._parse_function()))
4300
4301        if not expressions:
4302            self.raise_error("Failed to parse PIVOT's aggregation list")
4303
4304        if not self._match(TokenType.FOR):
4305            self.raise_error("Expecting FOR")
4306
4307        fields = []
4308        while True:
4309            field = self._try_parse(self._parse_pivot_in)
4310            if not field:
4311                break
4312            fields.append(field)
4313
4314        default_on_null = self._match_text_seq("DEFAULT", "ON", "NULL") and self._parse_wrapped(
4315            self._parse_bitwise
4316        )
4317
4318        group = self._parse_group()
4319
4320        self._match_r_paren()
4321
4322        pivot = self.expression(
4323            exp.Pivot,
4324            expressions=expressions,
4325            fields=fields,
4326            unpivot=unpivot,
4327            include_nulls=include_nulls,
4328            default_on_null=default_on_null,
4329            group=group,
4330        )
4331
4332        if not self._match_set((TokenType.PIVOT, TokenType.UNPIVOT), advance=False):
4333            pivot.set("alias", self._parse_table_alias())
4334
4335        if not unpivot:
4336            names = self._pivot_column_names(t.cast(t.List[exp.Expression], expressions))
4337
4338            columns: t.List[exp.Expression] = []
4339            all_fields = []
4340            for pivot_field in pivot.fields:
4341                pivot_field_expressions = pivot_field.expressions
4342
4343                # The `PivotAny` expression corresponds to `ANY ORDER BY <column>`; we can't infer in this case.
4344                if isinstance(seq_get(pivot_field_expressions, 0), exp.PivotAny):
4345                    continue
4346
4347                all_fields.append(
4348                    [
4349                        fld.sql() if self.IDENTIFY_PIVOT_STRINGS else fld.alias_or_name
4350                        for fld in pivot_field_expressions
4351                    ]
4352                )
4353
4354            if all_fields:
4355                if names:
4356                    all_fields.append(names)
4357
4358                # Generate all possible combinations of the pivot columns
4359                # e.g PIVOT(sum(...) as total FOR year IN (2000, 2010) FOR country IN ('NL', 'US'))
4360                # generates the product between [[2000, 2010], ['NL', 'US'], ['total']]
4361                for fld_parts_tuple in itertools.product(*all_fields):
4362                    fld_parts = list(fld_parts_tuple)
4363
4364                    if names and self.PREFIXED_PIVOT_COLUMNS:
4365                        # Move the "name" to the front of the list
4366                        fld_parts.insert(0, fld_parts.pop(-1))
4367
4368                    columns.append(exp.to_identifier("_".join(fld_parts)))
4369
4370            pivot.set("columns", columns)
4371
4372        return pivot
4373
4374    def _pivot_column_names(self, aggregations: t.List[exp.Expression]) -> t.List[str]:
4375        return [agg.alias for agg in aggregations if agg.alias]
4376
4377    def _parse_prewhere(self, skip_where_token: bool = False) -> t.Optional[exp.PreWhere]:
4378        if not skip_where_token and not self._match(TokenType.PREWHERE):
4379            return None
4380
4381        return self.expression(
4382            exp.PreWhere, comments=self._prev_comments, this=self._parse_assignment()
4383        )
4384
4385    def _parse_where(self, skip_where_token: bool = False) -> t.Optional[exp.Where]:
4386        if not skip_where_token and not self._match(TokenType.WHERE):
4387            return None
4388
4389        return self.expression(
4390            exp.Where, comments=self._prev_comments, this=self._parse_assignment()
4391        )
4392
4393    def _parse_group(self, skip_group_by_token: bool = False) -> t.Optional[exp.Group]:
4394        if not skip_group_by_token and not self._match(TokenType.GROUP_BY):
4395            return None
4396
4397        elements: t.Dict[str, t.Any] = defaultdict(list)
4398
4399        if self._match(TokenType.ALL):
4400            elements["all"] = True
4401        elif self._match(TokenType.DISTINCT):
4402            elements["all"] = False
4403
4404        while True:
4405            index = self._index
4406
4407            elements["expressions"].extend(
4408                self._parse_csv(
4409                    lambda: None
4410                    if self._match_set((TokenType.CUBE, TokenType.ROLLUP), advance=False)
4411                    else self._parse_assignment()
4412                )
4413            )
4414
4415            before_with_index = self._index
4416            with_prefix = self._match(TokenType.WITH)
4417
4418            if self._match(TokenType.ROLLUP):
4419                elements["rollup"].append(
4420                    self._parse_cube_or_rollup(exp.Rollup, with_prefix=with_prefix)
4421                )
4422            elif self._match(TokenType.CUBE):
4423                elements["cube"].append(
4424                    self._parse_cube_or_rollup(exp.Cube, with_prefix=with_prefix)
4425                )
4426            elif self._match(TokenType.GROUPING_SETS):
4427                elements["grouping_sets"].append(
4428                    self.expression(
4429                        exp.GroupingSets,
4430                        expressions=self._parse_wrapped_csv(self._parse_grouping_set),
4431                    )
4432                )
4433            elif self._match_text_seq("TOTALS"):
4434                elements["totals"] = True  # type: ignore
4435
4436            if before_with_index <= self._index <= before_with_index + 1:
4437                self._retreat(before_with_index)
4438                break
4439
4440            if index == self._index:
4441                break
4442
4443        return self.expression(exp.Group, **elements)  # type: ignore
4444
4445    def _parse_cube_or_rollup(self, kind: t.Type[E], with_prefix: bool = False) -> E:
4446        return self.expression(
4447            kind, expressions=[] if with_prefix else self._parse_wrapped_csv(self._parse_column)
4448        )
4449
4450    def _parse_grouping_set(self) -> t.Optional[exp.Expression]:
4451        if self._match(TokenType.L_PAREN):
4452            grouping_set = self._parse_csv(self._parse_column)
4453            self._match_r_paren()
4454            return self.expression(exp.Tuple, expressions=grouping_set)
4455
4456        return self._parse_column()
4457
4458    def _parse_having(self, skip_having_token: bool = False) -> t.Optional[exp.Having]:
4459        if not skip_having_token and not self._match(TokenType.HAVING):
4460            return None
4461        return self.expression(exp.Having, this=self._parse_assignment())
4462
4463    def _parse_qualify(self) -> t.Optional[exp.Qualify]:
4464        if not self._match(TokenType.QUALIFY):
4465            return None
4466        return self.expression(exp.Qualify, this=self._parse_assignment())
4467
4468    def _parse_connect_with_prior(self) -> t.Optional[exp.Expression]:
4469        self.NO_PAREN_FUNCTION_PARSERS["PRIOR"] = lambda self: self.expression(
4470            exp.Prior, this=self._parse_bitwise()
4471        )
4472        connect = self._parse_assignment()
4473        self.NO_PAREN_FUNCTION_PARSERS.pop("PRIOR")
4474        return connect
4475
4476    def _parse_connect(self, skip_start_token: bool = False) -> t.Optional[exp.Connect]:
4477        if skip_start_token:
4478            start = None
4479        elif self._match(TokenType.START_WITH):
4480            start = self._parse_assignment()
4481        else:
4482            return None
4483
4484        self._match(TokenType.CONNECT_BY)
4485        nocycle = self._match_text_seq("NOCYCLE")
4486        connect = self._parse_connect_with_prior()
4487
4488        if not start and self._match(TokenType.START_WITH):
4489            start = self._parse_assignment()
4490
4491        return self.expression(exp.Connect, start=start, connect=connect, nocycle=nocycle)
4492
4493    def _parse_name_as_expression(self) -> t.Optional[exp.Expression]:
4494        this = self._parse_id_var(any_token=True)
4495        if self._match(TokenType.ALIAS):
4496            this = self.expression(exp.Alias, alias=this, this=self._parse_assignment())
4497        return this
4498
4499    def _parse_interpolate(self) -> t.Optional[t.List[exp.Expression]]:
4500        if self._match_text_seq("INTERPOLATE"):
4501            return self._parse_wrapped_csv(self._parse_name_as_expression)
4502        return None
4503
4504    def _parse_order(
4505        self, this: t.Optional[exp.Expression] = None, skip_order_token: bool = False
4506    ) -> t.Optional[exp.Expression]:
4507        siblings = None
4508        if not skip_order_token and not self._match(TokenType.ORDER_BY):
4509            if not self._match(TokenType.ORDER_SIBLINGS_BY):
4510                return this
4511
4512            siblings = True
4513
4514        return self.expression(
4515            exp.Order,
4516            this=this,
4517            expressions=self._parse_csv(self._parse_ordered),
4518            siblings=siblings,
4519        )
4520
4521    def _parse_sort(self, exp_class: t.Type[E], token: TokenType) -> t.Optional[E]:
4522        if not self._match(token):
4523            return None
4524        return self.expression(exp_class, expressions=self._parse_csv(self._parse_ordered))
4525
4526    def _parse_ordered(
4527        self, parse_method: t.Optional[t.Callable] = None
4528    ) -> t.Optional[exp.Ordered]:
4529        this = parse_method() if parse_method else self._parse_assignment()
4530        if not this:
4531            return None
4532
4533        if this.name.upper() == "ALL" and self.dialect.SUPPORTS_ORDER_BY_ALL:
4534            this = exp.var("ALL")
4535
4536        asc = self._match(TokenType.ASC)
4537        desc = self._match(TokenType.DESC) or (asc and False)
4538
4539        is_nulls_first = self._match_text_seq("NULLS", "FIRST")
4540        is_nulls_last = self._match_text_seq("NULLS", "LAST")
4541
4542        nulls_first = is_nulls_first or False
4543        explicitly_null_ordered = is_nulls_first or is_nulls_last
4544
4545        if (
4546            not explicitly_null_ordered
4547            and (
4548                (not desc and self.dialect.NULL_ORDERING == "nulls_are_small")
4549                or (desc and self.dialect.NULL_ORDERING != "nulls_are_small")
4550            )
4551            and self.dialect.NULL_ORDERING != "nulls_are_last"
4552        ):
4553            nulls_first = True
4554
4555        if self._match_text_seq("WITH", "FILL"):
4556            with_fill = self.expression(
4557                exp.WithFill,
4558                **{  # type: ignore
4559                    "from": self._match(TokenType.FROM) and self._parse_bitwise(),
4560                    "to": self._match_text_seq("TO") and self._parse_bitwise(),
4561                    "step": self._match_text_seq("STEP") and self._parse_bitwise(),
4562                    "interpolate": self._parse_interpolate(),
4563                },
4564            )
4565        else:
4566            with_fill = None
4567
4568        return self.expression(
4569            exp.Ordered, this=this, desc=desc, nulls_first=nulls_first, with_fill=with_fill
4570        )
4571
4572    def _parse_limit_options(self) -> exp.LimitOptions:
4573        percent = self._match(TokenType.PERCENT)
4574        rows = self._match_set((TokenType.ROW, TokenType.ROWS))
4575        self._match_text_seq("ONLY")
4576        with_ties = self._match_text_seq("WITH", "TIES")
4577        return self.expression(exp.LimitOptions, percent=percent, rows=rows, with_ties=with_ties)
4578
4579    def _parse_limit(
4580        self,
4581        this: t.Optional[exp.Expression] = None,
4582        top: bool = False,
4583        skip_limit_token: bool = False,
4584    ) -> t.Optional[exp.Expression]:
4585        if skip_limit_token or self._match(TokenType.TOP if top else TokenType.LIMIT):
4586            comments = self._prev_comments
4587            if top:
4588                limit_paren = self._match(TokenType.L_PAREN)
4589                expression = self._parse_term() if limit_paren else self._parse_number()
4590
4591                if limit_paren:
4592                    self._match_r_paren()
4593
4594                limit_options = self._parse_limit_options()
4595            else:
4596                limit_options = None
4597                expression = self._parse_term()
4598
4599            if self._match(TokenType.COMMA):
4600                offset = expression
4601                expression = self._parse_term()
4602            else:
4603                offset = None
4604
4605            limit_exp = self.expression(
4606                exp.Limit,
4607                this=this,
4608                expression=expression,
4609                offset=offset,
4610                comments=comments,
4611                limit_options=limit_options,
4612                expressions=self._parse_limit_by(),
4613            )
4614
4615            return limit_exp
4616
4617        if self._match(TokenType.FETCH):
4618            direction = self._match_set((TokenType.FIRST, TokenType.NEXT))
4619            direction = self._prev.text.upper() if direction else "FIRST"
4620
4621            count = self._parse_field(tokens=self.FETCH_TOKENS)
4622
4623            return self.expression(
4624                exp.Fetch,
4625                direction=direction,
4626                count=count,
4627                limit_options=self._parse_limit_options(),
4628            )
4629
4630        return this
4631
4632    def _parse_offset(self, this: t.Optional[exp.Expression] = None) -> t.Optional[exp.Expression]:
4633        if not self._match(TokenType.OFFSET):
4634            return this
4635
4636        count = self._parse_term()
4637        self._match_set((TokenType.ROW, TokenType.ROWS))
4638
4639        return self.expression(
4640            exp.Offset, this=this, expression=count, expressions=self._parse_limit_by()
4641        )
4642
4643    def _can_parse_limit_or_offset(self) -> bool:
4644        if not self._match_set(self.AMBIGUOUS_ALIAS_TOKENS, advance=False):
4645            return False
4646
4647        index = self._index
4648        result = bool(
4649            self._try_parse(self._parse_limit, retreat=True)
4650            or self._try_parse(self._parse_offset, retreat=True)
4651        )
4652        self._retreat(index)
4653        return result
4654
4655    def _parse_limit_by(self) -> t.Optional[t.List[exp.Expression]]:
4656        return self._match_text_seq("BY") and self._parse_csv(self._parse_bitwise)
4657
4658    def _parse_locks(self) -> t.List[exp.Lock]:
4659        locks = []
4660        while True:
4661            if self._match_text_seq("FOR", "UPDATE"):
4662                update = True
4663            elif self._match_text_seq("FOR", "SHARE") or self._match_text_seq(
4664                "LOCK", "IN", "SHARE", "MODE"
4665            ):
4666                update = False
4667            else:
4668                break
4669
4670            expressions = None
4671            if self._match_text_seq("OF"):
4672                expressions = self._parse_csv(lambda: self._parse_table(schema=True))
4673
4674            wait: t.Optional[bool | exp.Expression] = None
4675            if self._match_text_seq("NOWAIT"):
4676                wait = True
4677            elif self._match_text_seq("WAIT"):
4678                wait = self._parse_primary()
4679            elif self._match_text_seq("SKIP", "LOCKED"):
4680                wait = False
4681
4682            locks.append(
4683                self.expression(exp.Lock, update=update, expressions=expressions, wait=wait)
4684            )
4685
4686        return locks
4687
4688    def parse_set_operation(self, this: t.Optional[exp.Expression]) -> t.Optional[exp.Expression]:
4689        start = self._index
4690        _, side_token, kind_token = self._parse_join_parts()
4691
4692        side = side_token.text if side_token else None
4693        kind = kind_token.text if kind_token else None
4694
4695        if not self._match_set(self.SET_OPERATIONS):
4696            self._retreat(start)
4697            return None
4698
4699        token_type = self._prev.token_type
4700
4701        if token_type == TokenType.UNION:
4702            operation: t.Type[exp.SetOperation] = exp.Union
4703        elif token_type == TokenType.EXCEPT:
4704            operation = exp.Except
4705        else:
4706            operation = exp.Intersect
4707
4708        comments = self._prev.comments
4709
4710        if self._match(TokenType.DISTINCT):
4711            distinct: t.Optional[bool] = True
4712        elif self._match(TokenType.ALL):
4713            distinct = False
4714        else:
4715            distinct = self.dialect.SET_OP_DISTINCT_BY_DEFAULT[operation]
4716            if distinct is None:
4717                self.raise_error(f"Expected DISTINCT or ALL for {operation.__name__}")
4718
4719        by_name = self._match_text_seq("BY", "NAME") or self._match_text_seq(
4720            "STRICT", "CORRESPONDING"
4721        )
4722        if self._match_text_seq("CORRESPONDING"):
4723            by_name = True
4724            if not side and not kind:
4725                kind = "INNER"
4726
4727        on_column_list = None
4728        if by_name and self._match_texts(("ON", "BY")):
4729            on_column_list = self._parse_wrapped_csv(self._parse_column)
4730
4731        expression = self._parse_select(nested=True, parse_set_operation=False)
4732
4733        return self.expression(
4734            operation,
4735            comments=comments,
4736            this=this,
4737            distinct=distinct,
4738            by_name=by_name,
4739            expression=expression,
4740            side=side,
4741            kind=kind,
4742            on=on_column_list,
4743        )
4744
4745    def _parse_set_operations(self, this: t.Optional[exp.Expression]) -> t.Optional[exp.Expression]:
4746        while this:
4747            setop = self.parse_set_operation(this)
4748            if not setop:
4749                break
4750            this = setop
4751
4752        if isinstance(this, exp.SetOperation) and self.MODIFIERS_ATTACHED_TO_SET_OP:
4753            expression = this.expression
4754
4755            if expression:
4756                for arg in self.SET_OP_MODIFIERS:
4757                    expr = expression.args.get(arg)
4758                    if expr:
4759                        this.set(arg, expr.pop())
4760
4761        return this
4762
4763    def _parse_expression(self) -> t.Optional[exp.Expression]:
4764        return self._parse_alias(self._parse_assignment())
4765
4766    def _parse_assignment(self) -> t.Optional[exp.Expression]:
4767        this = self._parse_disjunction()
4768        if not this and self._next and self._next.token_type in self.ASSIGNMENT:
4769            # This allows us to parse <non-identifier token> := <expr>
4770            this = exp.column(
4771                t.cast(str, self._advance_any(ignore_reserved=True) and self._prev.text)
4772            )
4773
4774        while self._match_set(self.ASSIGNMENT):
4775            if isinstance(this, exp.Column) and len(this.parts) == 1:
4776                this = this.this
4777
4778            this = self.expression(
4779                self.ASSIGNMENT[self._prev.token_type],
4780                this=this,
4781                comments=self._prev_comments,
4782                expression=self._parse_assignment(),
4783            )
4784
4785        return this
4786
4787    def _parse_disjunction(self) -> t.Optional[exp.Expression]:
4788        return self._parse_tokens(self._parse_conjunction, self.DISJUNCTION)
4789
4790    def _parse_conjunction(self) -> t.Optional[exp.Expression]:
4791        return self._parse_tokens(self._parse_equality, self.CONJUNCTION)
4792
4793    def _parse_equality(self) -> t.Optional[exp.Expression]:
4794        return self._parse_tokens(self._parse_comparison, self.EQUALITY)
4795
4796    def _parse_comparison(self) -> t.Optional[exp.Expression]:
4797        return self._parse_tokens(self._parse_range, self.COMPARISON)
4798
4799    def _parse_range(self, this: t.Optional[exp.Expression] = None) -> t.Optional[exp.Expression]:
4800        this = this or self._parse_bitwise()
4801        negate = self._match(TokenType.NOT)
4802
4803        if self._match_set(self.RANGE_PARSERS):
4804            expression = self.RANGE_PARSERS[self._prev.token_type](self, this)
4805            if not expression:
4806                return this
4807
4808            this = expression
4809        elif self._match(TokenType.ISNULL):
4810            this = self.expression(exp.Is, this=this, expression=exp.Null())
4811
4812        # Postgres supports ISNULL and NOTNULL for conditions.
4813        # https://blog.andreiavram.ro/postgresql-null-composite-type/
4814        if self._match(TokenType.NOTNULL):
4815            this = self.expression(exp.Is, this=this, expression=exp.Null())
4816            this = self.expression(exp.Not, this=this)
4817
4818        if negate:
4819            this = self._negate_range(this)
4820
4821        if self._match(TokenType.IS):
4822            this = self._parse_is(this)
4823
4824        return this
4825
4826    def _negate_range(self, this: t.Optional[exp.Expression] = None) -> t.Optional[exp.Expression]:
4827        if not this:
4828            return this
4829
4830        return self.expression(exp.Not, this=this)
4831
4832    def _parse_is(self, this: t.Optional[exp.Expression]) -> t.Optional[exp.Expression]:
4833        index = self._index - 1
4834        negate = self._match(TokenType.NOT)
4835
4836        if self._match_text_seq("DISTINCT", "FROM"):
4837            klass = exp.NullSafeEQ if negate else exp.NullSafeNEQ
4838            return self.expression(klass, this=this, expression=self._parse_bitwise())
4839
4840        if self._match(TokenType.JSON):
4841            kind = self._match_texts(self.IS_JSON_PREDICATE_KIND) and self._prev.text.upper()
4842
4843            if self._match_text_seq("WITH"):
4844                _with = True
4845            elif self._match_text_seq("WITHOUT"):
4846                _with = False
4847            else:
4848                _with = None
4849
4850            unique = self._match(TokenType.UNIQUE)
4851            self._match_text_seq("KEYS")
4852            expression: t.Optional[exp.Expression] = self.expression(
4853                exp.JSON, **{"this": kind, "with": _with, "unique": unique}
4854            )
4855        else:
4856            expression = self._parse_primary() or self._parse_null()
4857            if not expression:
4858                self._retreat(index)
4859                return None
4860
4861        this = self.expression(exp.Is, this=this, expression=expression)
4862        return self.expression(exp.Not, this=this) if negate else this
4863
4864    def _parse_in(self, this: t.Optional[exp.Expression], alias: bool = False) -> exp.In:
4865        unnest = self._parse_unnest(with_alias=False)
4866        if unnest:
4867            this = self.expression(exp.In, this=this, unnest=unnest)
4868        elif self._match_set((TokenType.L_PAREN, TokenType.L_BRACKET)):
4869            matched_l_paren = self._prev.token_type == TokenType.L_PAREN
4870            expressions = self._parse_csv(lambda: self._parse_select_or_expression(alias=alias))
4871
4872            if len(expressions) == 1 and isinstance(expressions[0], exp.Query):
4873                this = self.expression(exp.In, this=this, query=expressions[0].subquery(copy=False))
4874            else:
4875                this = self.expression(exp.In, this=this, expressions=expressions)
4876
4877            if matched_l_paren:
4878                self._match_r_paren(this)
4879            elif not self._match(TokenType.R_BRACKET, expression=this):
4880                self.raise_error("Expecting ]")
4881        else:
4882            this = self.expression(exp.In, this=this, field=self._parse_column())
4883
4884        return this
4885
4886    def _parse_between(self, this: t.Optional[exp.Expression]) -> exp.Between:
4887        low = self._parse_bitwise()
4888        self._match(TokenType.AND)
4889        high = self._parse_bitwise()
4890        return self.expression(exp.Between, this=this, low=low, high=high)
4891
4892    def _parse_escape(self, this: t.Optional[exp.Expression]) -> t.Optional[exp.Expression]:
4893        if not self._match(TokenType.ESCAPE):
4894            return this
4895        return self.expression(exp.Escape, this=this, expression=self._parse_string())
4896
4897    def _parse_interval(self, match_interval: bool = True) -> t.Optional[exp.Add | exp.Interval]:
4898        index = self._index
4899
4900        if not self._match(TokenType.INTERVAL) and match_interval:
4901            return None
4902
4903        if self._match(TokenType.STRING, advance=False):
4904            this = self._parse_primary()
4905        else:
4906            this = self._parse_term()
4907
4908        if not this or (
4909            isinstance(this, exp.Column)
4910            and not this.table
4911            and not this.this.quoted
4912            and this.name.upper() == "IS"
4913        ):
4914            self._retreat(index)
4915            return None
4916
4917        unit = self._parse_function() or (
4918            not self._match(TokenType.ALIAS, advance=False)
4919            and self._parse_var(any_token=True, upper=True)
4920        )
4921
4922        # Most dialects support, e.g., the form INTERVAL '5' day, thus we try to parse
4923        # each INTERVAL expression into this canonical form so it's easy to transpile
4924        if this and this.is_number:
4925            this = exp.Literal.string(this.to_py())
4926        elif this and this.is_string:
4927            parts = exp.INTERVAL_STRING_RE.findall(this.name)
4928            if parts and unit:
4929                # Unconsume the eagerly-parsed unit, since the real unit was part of the string
4930                unit = None
4931                self._retreat(self._index - 1)
4932
4933            if len(parts) == 1:
4934                this = exp.Literal.string(parts[0][0])
4935                unit = self.expression(exp.Var, this=parts[0][1].upper())
4936        if self.INTERVAL_SPANS and self._match_text_seq("TO"):
4937            unit = self.expression(
4938                exp.IntervalSpan, this=unit, expression=self._parse_var(any_token=True, upper=True)
4939            )
4940
4941        interval = self.expression(exp.Interval, this=this, unit=unit)
4942
4943        index = self._index
4944        self._match(TokenType.PLUS)
4945
4946        # Convert INTERVAL 'val_1' unit_1 [+] ... [+] 'val_n' unit_n into a sum of intervals
4947        if self._match_set((TokenType.STRING, TokenType.NUMBER), advance=False):
4948            return self.expression(
4949                exp.Add, this=interval, expression=self._parse_interval(match_interval=False)
4950            )
4951
4952        self._retreat(index)
4953        return interval
4954
4955    def _parse_bitwise(self) -> t.Optional[exp.Expression]:
4956        this = self._parse_term()
4957
4958        while True:
4959            if self._match_set(self.BITWISE):
4960                this = self.expression(
4961                    self.BITWISE[self._prev.token_type],
4962                    this=this,
4963                    expression=self._parse_term(),
4964                )
4965            elif self.dialect.DPIPE_IS_STRING_CONCAT and self._match(TokenType.DPIPE):
4966                this = self.expression(
4967                    exp.DPipe,
4968                    this=this,
4969                    expression=self._parse_term(),
4970                    safe=not self.dialect.STRICT_STRING_CONCAT,
4971                )
4972            elif self._match(TokenType.DQMARK):
4973                this = self.expression(
4974                    exp.Coalesce, this=this, expressions=ensure_list(self._parse_term())
4975                )
4976            elif self._match_pair(TokenType.LT, TokenType.LT):
4977                this = self.expression(
4978                    exp.BitwiseLeftShift, this=this, expression=self._parse_term()
4979                )
4980            elif self._match_pair(TokenType.GT, TokenType.GT):
4981                this = self.expression(
4982                    exp.BitwiseRightShift, this=this, expression=self._parse_term()
4983                )
4984            else:
4985                break
4986
4987        return this
4988
4989    def _parse_term(self) -> t.Optional[exp.Expression]:
4990        this = self._parse_factor()
4991
4992        while self._match_set(self.TERM):
4993            klass = self.TERM[self._prev.token_type]
4994            comments = self._prev_comments
4995            expression = self._parse_factor()
4996
4997            this = self.expression(klass, this=this, comments=comments, expression=expression)
4998
4999            if isinstance(this, exp.Collate):
5000                expr = this.expression
5001
5002                # Preserve collations such as pg_catalog."default" (Postgres) as columns, otherwise
5003                # fallback to Identifier / Var
5004                if isinstance(expr, exp.Column) and len(expr.parts) == 1:
5005                    ident = expr.this
5006                    if isinstance(ident, exp.Identifier):
5007                        this.set("expression", ident if ident.quoted else exp.var(ident.name))
5008
5009        return this
5010
5011    def _parse_factor(self) -> t.Optional[exp.Expression]:
5012        parse_method = self._parse_exponent if self.EXPONENT else self._parse_unary
5013        this = parse_method()
5014
5015        while self._match_set(self.FACTOR):
5016            klass = self.FACTOR[self._prev.token_type]
5017            comments = self._prev_comments
5018            expression = parse_method()
5019
5020            if not expression and klass is exp.IntDiv and self._prev.text.isalpha():
5021                self._retreat(self._index - 1)
5022                return this
5023
5024            this = self.expression(klass, this=this, comments=comments, expression=expression)
5025
5026            if isinstance(this, exp.Div):
5027                this.args["typed"] = self.dialect.TYPED_DIVISION
5028                this.args["safe"] = self.dialect.SAFE_DIVISION
5029
5030        return this
5031
5032    def _parse_exponent(self) -> t.Optional[exp.Expression]:
5033        return self._parse_tokens(self._parse_unary, self.EXPONENT)
5034
5035    def _parse_unary(self) -> t.Optional[exp.Expression]:
5036        if self._match_set(self.UNARY_PARSERS):
5037            return self.UNARY_PARSERS[self._prev.token_type](self)
5038        return self._parse_at_time_zone(self._parse_type())
5039
5040    def _parse_type(
5041        self, parse_interval: bool = True, fallback_to_identifier: bool = False
5042    ) -> t.Optional[exp.Expression]:
5043        interval = parse_interval and self._parse_interval()
5044        if interval:
5045            return interval
5046
5047        index = self._index
5048        data_type = self._parse_types(check_func=True, allow_identifiers=False)
5049
5050        # parse_types() returns a Cast if we parsed BQ's inline constructor <type>(<values>) e.g.
5051        # STRUCT<a INT, b STRING>(1, 'foo'), which is canonicalized to CAST(<values> AS <type>)
5052        if isinstance(data_type, exp.Cast):
5053            # This constructor can contain ops directly after it, for instance struct unnesting:
5054            # STRUCT<a INT, b STRING>(1, 'foo').* --> CAST(STRUCT(1, 'foo') AS STRUCT<a iNT, b STRING).*
5055            return self._parse_column_ops(data_type)
5056
5057        if data_type:
5058            index2 = self._index
5059            this = self._parse_primary()
5060
5061            if isinstance(this, exp.Literal):
5062                this = self._parse_column_ops(this)
5063
5064                parser = self.TYPE_LITERAL_PARSERS.get(data_type.this)
5065                if parser:
5066                    return parser(self, this, data_type)
5067
5068                return self.expression(exp.Cast, this=this, to=data_type)
5069
5070            # The expressions arg gets set by the parser when we have something like DECIMAL(38, 0)
5071            # in the input SQL. In that case, we'll produce these tokens: DECIMAL ( 38 , 0 )
5072            #
5073            # If the index difference here is greater than 1, that means the parser itself must have
5074            # consumed additional tokens such as the DECIMAL scale and precision in the above example.
5075            #
5076            # If it's not greater than 1, then it must be 1, because we've consumed at least the type
5077            # keyword, meaning that the expressions arg of the DataType must have gotten set by a
5078            # callable in the TYPE_CONVERTERS mapping. For example, Snowflake converts DECIMAL to
5079            # DECIMAL(38, 0)) in order to facilitate the data type's transpilation.
5080            #
5081            # In these cases, we don't really want to return the converted type, but instead retreat
5082            # and try to parse a Column or Identifier in the section below.
5083            if data_type.expressions and index2 - index > 1:
5084                self._retreat(index2)
5085                return self._parse_column_ops(data_type)
5086
5087            self._retreat(index)
5088
5089        if fallback_to_identifier:
5090            return self._parse_id_var()
5091
5092        this = self._parse_column()
5093        return this and self._parse_column_ops(this)
5094
5095    def _parse_type_size(self) -> t.Optional[exp.DataTypeParam]:
5096        this = self._parse_type()
5097        if not this:
5098            return None
5099
5100        if isinstance(this, exp.Column) and not this.table:
5101            this = exp.var(this.name.upper())
5102
5103        return self.expression(
5104            exp.DataTypeParam, this=this, expression=self._parse_var(any_token=True)
5105        )
5106
5107    def _parse_types(
5108        self, check_func: bool = False, schema: bool = False, allow_identifiers: bool = True
5109    ) -> t.Optional[exp.Expression]:
5110        index = self._index
5111
5112        this: t.Optional[exp.Expression] = None
5113        prefix = self._match_text_seq("SYSUDTLIB", ".")
5114
5115        if not self._match_set(self.TYPE_TOKENS):
5116            identifier = allow_identifiers and self._parse_id_var(
5117                any_token=False, tokens=(TokenType.VAR,)
5118            )
5119            if isinstance(identifier, exp.Identifier):
5120                tokens = self.dialect.tokenize(identifier.sql(dialect=self.dialect))
5121
5122                if len(tokens) != 1:
5123                    self.raise_error("Unexpected identifier", self._prev)
5124
5125                if tokens[0].token_type in self.TYPE_TOKENS:
5126                    self._prev = tokens[0]
5127                elif self.dialect.SUPPORTS_USER_DEFINED_TYPES:
5128                    type_name = identifier.name
5129
5130                    while self._match(TokenType.DOT):
5131                        type_name = f"{type_name}.{self._advance_any() and self._prev.text}"
5132
5133                    this = exp.DataType.build(type_name, udt=True)
5134                else:
5135                    self._retreat(self._index - 1)
5136                    return None
5137            else:
5138                return None
5139
5140        type_token = self._prev.token_type
5141
5142        if type_token == TokenType.PSEUDO_TYPE:
5143            return self.expression(exp.PseudoType, this=self._prev.text.upper())
5144
5145        if type_token == TokenType.OBJECT_IDENTIFIER:
5146            return self.expression(exp.ObjectIdentifier, this=self._prev.text.upper())
5147
5148        # https://materialize.com/docs/sql/types/map/
5149        if type_token == TokenType.MAP and self._match(TokenType.L_BRACKET):
5150            key_type = self._parse_types(
5151                check_func=check_func, schema=schema, allow_identifiers=allow_identifiers
5152            )
5153            if not self._match(TokenType.FARROW):
5154                self._retreat(index)
5155                return None
5156
5157            value_type = self._parse_types(
5158                check_func=check_func, schema=schema, allow_identifiers=allow_identifiers
5159            )
5160            if not self._match(TokenType.R_BRACKET):
5161                self._retreat(index)
5162                return None
5163
5164            return exp.DataType(
5165                this=exp.DataType.Type.MAP,
5166                expressions=[key_type, value_type],
5167                nested=True,
5168                prefix=prefix,
5169            )
5170
5171        nested = type_token in self.NESTED_TYPE_TOKENS
5172        is_struct = type_token in self.STRUCT_TYPE_TOKENS
5173        is_aggregate = type_token in self.AGGREGATE_TYPE_TOKENS
5174        expressions = None
5175        maybe_func = False
5176
5177        if self._match(TokenType.L_PAREN):
5178            if is_struct:
5179                expressions = self._parse_csv(lambda: self._parse_struct_types(type_required=True))
5180            elif nested:
5181                expressions = self._parse_csv(
5182                    lambda: self._parse_types(
5183                        check_func=check_func, schema=schema, allow_identifiers=allow_identifiers
5184                    )
5185                )
5186                if type_token == TokenType.NULLABLE and len(expressions) == 1:
5187                    this = expressions[0]
5188                    this.set("nullable", True)
5189                    self._match_r_paren()
5190                    return this
5191            elif type_token in self.ENUM_TYPE_TOKENS:
5192                expressions = self._parse_csv(self._parse_equality)
5193            elif is_aggregate:
5194                func_or_ident = self._parse_function(anonymous=True) or self._parse_id_var(
5195                    any_token=False, tokens=(TokenType.VAR, TokenType.ANY)
5196                )
5197                if not func_or_ident:
5198                    return None
5199                expressions = [func_or_ident]
5200                if self._match(TokenType.COMMA):
5201                    expressions.extend(
5202                        self._parse_csv(
5203                            lambda: self._parse_types(
5204                                check_func=check_func,
5205                                schema=schema,
5206                                allow_identifiers=allow_identifiers,
5207                            )
5208                        )
5209                    )
5210            else:
5211                expressions = self._parse_csv(self._parse_type_size)
5212
5213                # https://docs.snowflake.com/en/sql-reference/data-types-vector
5214                if type_token == TokenType.VECTOR and len(expressions) == 2:
5215                    expressions[0] = exp.DataType.build(expressions[0].name, dialect=self.dialect)
5216
5217            if not expressions or not self._match(TokenType.R_PAREN):
5218                self._retreat(index)
5219                return None
5220
5221            maybe_func = True
5222
5223        values: t.Optional[t.List[exp.Expression]] = None
5224
5225        if nested and self._match(TokenType.LT):
5226            if is_struct:
5227                expressions = self._parse_csv(lambda: self._parse_struct_types(type_required=True))
5228            else:
5229                expressions = self._parse_csv(
5230                    lambda: self._parse_types(
5231                        check_func=check_func, schema=schema, allow_identifiers=allow_identifiers
5232                    )
5233                )
5234
5235            if not self._match(TokenType.GT):
5236                self.raise_error("Expecting >")
5237
5238            if self._match_set((TokenType.L_BRACKET, TokenType.L_PAREN)):
5239                values = self._parse_csv(self._parse_assignment)
5240                if not values and is_struct:
5241                    values = None
5242                    self._retreat(self._index - 1)
5243                else:
5244                    self._match_set((TokenType.R_BRACKET, TokenType.R_PAREN))
5245
5246        if type_token in self.TIMESTAMPS:
5247            if self._match_text_seq("WITH", "TIME", "ZONE"):
5248                maybe_func = False
5249                tz_type = (
5250                    exp.DataType.Type.TIMETZ
5251                    if type_token in self.TIMES
5252                    else exp.DataType.Type.TIMESTAMPTZ
5253                )
5254                this = exp.DataType(this=tz_type, expressions=expressions)
5255            elif self._match_text_seq("WITH", "LOCAL", "TIME", "ZONE"):
5256                maybe_func = False
5257                this = exp.DataType(this=exp.DataType.Type.TIMESTAMPLTZ, expressions=expressions)
5258            elif self._match_text_seq("WITHOUT", "TIME", "ZONE"):
5259                maybe_func = False
5260        elif type_token == TokenType.INTERVAL:
5261            unit = self._parse_var(upper=True)
5262            if unit:
5263                if self._match_text_seq("TO"):
5264                    unit = exp.IntervalSpan(this=unit, expression=self._parse_var(upper=True))
5265
5266                this = self.expression(exp.DataType, this=self.expression(exp.Interval, unit=unit))
5267            else:
5268                this = self.expression(exp.DataType, this=exp.DataType.Type.INTERVAL)
5269        elif type_token == TokenType.VOID:
5270            this = exp.DataType(this=exp.DataType.Type.NULL)
5271
5272        if maybe_func and check_func:
5273            index2 = self._index
5274            peek = self._parse_string()
5275
5276            if not peek:
5277                self._retreat(index)
5278                return None
5279
5280            self._retreat(index2)
5281
5282        if not this:
5283            if self._match_text_seq("UNSIGNED"):
5284                unsigned_type_token = self.SIGNED_TO_UNSIGNED_TYPE_TOKEN.get(type_token)
5285                if not unsigned_type_token:
5286                    self.raise_error(f"Cannot convert {type_token.value} to unsigned.")
5287
5288                type_token = unsigned_type_token or type_token
5289
5290            this = exp.DataType(
5291                this=exp.DataType.Type[type_token.value],
5292                expressions=expressions,
5293                nested=nested,
5294                prefix=prefix,
5295            )
5296
5297            # Empty arrays/structs are allowed
5298            if values is not None:
5299                cls = exp.Struct if is_struct else exp.Array
5300                this = exp.cast(cls(expressions=values), this, copy=False)
5301
5302        elif expressions:
5303            this.set("expressions", expressions)
5304
5305        # https://materialize.com/docs/sql/types/list/#type-name
5306        while self._match(TokenType.LIST):
5307            this = exp.DataType(this=exp.DataType.Type.LIST, expressions=[this], nested=True)
5308
5309        index = self._index
5310
5311        # Postgres supports the INT ARRAY[3] syntax as a synonym for INT[3]
5312        matched_array = self._match(TokenType.ARRAY)
5313
5314        while self._curr:
5315            datatype_token = self._prev.token_type
5316            matched_l_bracket = self._match(TokenType.L_BRACKET)
5317
5318            if (not matched_l_bracket and not matched_array) or (
5319                datatype_token == TokenType.ARRAY and self._match(TokenType.R_BRACKET)
5320            ):
5321                # Postgres allows casting empty arrays such as ARRAY[]::INT[],
5322                # not to be confused with the fixed size array parsing
5323                break
5324
5325            matched_array = False
5326            values = self._parse_csv(self._parse_assignment) or None
5327            if (
5328                values
5329                and not schema
5330                and (
5331                    not self.dialect.SUPPORTS_FIXED_SIZE_ARRAYS or datatype_token == TokenType.ARRAY
5332                )
5333            ):
5334                # Retreating here means that we should not parse the following values as part of the data type, e.g. in DuckDB
5335                # ARRAY[1] should retreat and instead be parsed into exp.Array in contrast to INT[x][y] which denotes a fixed-size array data type
5336                self._retreat(index)
5337                break
5338
5339            this = exp.DataType(
5340                this=exp.DataType.Type.ARRAY, expressions=[this], values=values, nested=True
5341            )
5342            self._match(TokenType.R_BRACKET)
5343
5344        if self.TYPE_CONVERTERS and isinstance(this.this, exp.DataType.Type):
5345            converter = self.TYPE_CONVERTERS.get(this.this)
5346            if converter:
5347                this = converter(t.cast(exp.DataType, this))
5348
5349        return this
5350
5351    def _parse_struct_types(self, type_required: bool = False) -> t.Optional[exp.Expression]:
5352        index = self._index
5353
5354        if (
5355            self._curr
5356            and self._next
5357            and self._curr.token_type in self.TYPE_TOKENS
5358            and self._next.token_type in self.TYPE_TOKENS
5359        ):
5360            # Takes care of special cases like `STRUCT<list ARRAY<...>>` where the identifier is also a
5361            # type token. Without this, the list will be parsed as a type and we'll eventually crash
5362            this = self._parse_id_var()
5363        else:
5364            this = (
5365                self._parse_type(parse_interval=False, fallback_to_identifier=True)
5366                or self._parse_id_var()
5367            )
5368
5369        self._match(TokenType.COLON)
5370
5371        if (
5372            type_required
5373            and not isinstance(this, exp.DataType)
5374            and not self._match_set(self.TYPE_TOKENS, advance=False)
5375        ):
5376            self._retreat(index)
5377            return self._parse_types()
5378
5379        return self._parse_column_def(this)
5380
5381    def _parse_at_time_zone(self, this: t.Optional[exp.Expression]) -> t.Optional[exp.Expression]:
5382        if not self._match_text_seq("AT", "TIME", "ZONE"):
5383            return this
5384        return self.expression(exp.AtTimeZone, this=this, zone=self._parse_unary())
5385
5386    def _parse_column(self) -> t.Optional[exp.Expression]:
5387        this = self._parse_column_reference()
5388        column = self._parse_column_ops(this) if this else self._parse_bracket(this)
5389
5390        if self.dialect.SUPPORTS_COLUMN_JOIN_MARKS and column:
5391            column.set("join_mark", self._match(TokenType.JOIN_MARKER))
5392
5393        return column
5394
5395    def _parse_column_reference(self) -> t.Optional[exp.Expression]:
5396        this = self._parse_field()
5397        if (
5398            not this
5399            and self._match(TokenType.VALUES, advance=False)
5400            and self.VALUES_FOLLOWED_BY_PAREN
5401            and (not self._next or self._next.token_type != TokenType.L_PAREN)
5402        ):
5403            this = self._parse_id_var()
5404
5405        if isinstance(this, exp.Identifier):
5406            # We bubble up comments from the Identifier to the Column
5407            this = self.expression(exp.Column, comments=this.pop_comments(), this=this)
5408
5409        return this
5410
5411    def _parse_colon_as_variant_extract(
5412        self, this: t.Optional[exp.Expression]
5413    ) -> t.Optional[exp.Expression]:
5414        casts = []
5415        json_path = []
5416        escape = None
5417
5418        while self._match(TokenType.COLON):
5419            start_index = self._index
5420
5421            # Snowflake allows reserved keywords as json keys but advance_any() excludes TokenType.SELECT from any_tokens=True
5422            path = self._parse_column_ops(
5423                self._parse_field(any_token=True, tokens=(TokenType.SELECT,))
5424            )
5425
5426            # The cast :: operator has a lower precedence than the extraction operator :, so
5427            # we rearrange the AST appropriately to avoid casting the JSON path
5428            while isinstance(path, exp.Cast):
5429                casts.append(path.to)
5430                path = path.this
5431
5432            if casts:
5433                dcolon_offset = next(
5434                    i
5435                    for i, t in enumerate(self._tokens[start_index:])
5436                    if t.token_type == TokenType.DCOLON
5437                )
5438                end_token = self._tokens[start_index + dcolon_offset - 1]
5439            else:
5440                end_token = self._prev
5441
5442            if path:
5443                # Escape single quotes from Snowflake's colon extraction (e.g. col:"a'b") as
5444                # it'll roundtrip to a string literal in GET_PATH
5445                if isinstance(path, exp.Identifier) and path.quoted:
5446                    escape = True
5447
5448                json_path.append(self._find_sql(self._tokens[start_index], end_token))
5449
5450        # The VARIANT extract in Snowflake/Databricks is parsed as a JSONExtract; Snowflake uses the json_path in GET_PATH() while
5451        # Databricks transforms it back to the colon/dot notation
5452        if json_path:
5453            json_path_expr = self.dialect.to_json_path(exp.Literal.string(".".join(json_path)))
5454
5455            if json_path_expr:
5456                json_path_expr.set("escape", escape)
5457
5458            this = self.expression(
5459                exp.JSONExtract,
5460                this=this,
5461                expression=json_path_expr,
5462                variant_extract=True,
5463            )
5464
5465            while casts:
5466                this = self.expression(exp.Cast, this=this, to=casts.pop())
5467
5468        return this
5469
5470    def _parse_dcolon(self) -> t.Optional[exp.Expression]:
5471        return self._parse_types()
5472
5473    def _parse_column_ops(self, this: t.Optional[exp.Expression]) -> t.Optional[exp.Expression]:
5474        this = self._parse_bracket(this)
5475
5476        while self._match_set(self.COLUMN_OPERATORS):
5477            op_token = self._prev.token_type
5478            op = self.COLUMN_OPERATORS.get(op_token)
5479
5480            if op_token in (TokenType.DCOLON, TokenType.DOTCOLON):
5481                field = self._parse_dcolon()
5482                if not field:
5483                    self.raise_error("Expected type")
5484            elif op and self._curr:
5485                field = self._parse_column_reference() or self._parse_bracket()
5486                if isinstance(field, exp.Column) and self._match(TokenType.DOT, advance=False):
5487                    field = self._parse_column_ops(field)
5488            else:
5489                field = self._parse_field(any_token=True, anonymous_func=True)
5490
5491            if isinstance(field, (exp.Func, exp.Window)) and this:
5492                # BQ & snowflake allow function calls like x.y.count(...), SAFE.SUBSTR(...) etc
5493                # https://cloud.google.com/bigquery/docs/reference/standard-sql/functions-reference#function_call_rules
5494                this = exp.replace_tree(
5495                    this,
5496                    lambda n: (
5497                        self.expression(exp.Dot, this=n.args.get("table"), expression=n.this)
5498                        if n.table
5499                        else n.this
5500                    )
5501                    if isinstance(n, exp.Column)
5502                    else n,
5503                )
5504
5505            if op:
5506                this = op(self, this, field)
5507            elif isinstance(this, exp.Column) and not this.args.get("catalog"):
5508                this = self.expression(
5509                    exp.Column,
5510                    comments=this.comments,
5511                    this=field,
5512                    table=this.this,
5513                    db=this.args.get("table"),
5514                    catalog=this.args.get("db"),
5515                )
5516            elif isinstance(field, exp.Window):
5517                # Move the exp.Dot's to the window's function
5518                window_func = self.expression(exp.Dot, this=this, expression=field.this)
5519                field.set("this", window_func)
5520                this = field
5521            else:
5522                this = self.expression(exp.Dot, this=this, expression=field)
5523
5524            if field and field.comments:
5525                t.cast(exp.Expression, this).add_comments(field.pop_comments())
5526
5527            this = self._parse_bracket(this)
5528
5529        return self._parse_colon_as_variant_extract(this) if self.COLON_IS_VARIANT_EXTRACT else this
5530
5531    def _parse_primary(self) -> t.Optional[exp.Expression]:
5532        if self._match_set(self.PRIMARY_PARSERS):
5533            token_type = self._prev.token_type
5534            primary = self.PRIMARY_PARSERS[token_type](self, self._prev)
5535
5536            if token_type == TokenType.STRING:
5537                expressions = [primary]
5538                while self._match(TokenType.STRING):
5539                    expressions.append(exp.Literal.string(self._prev.text))
5540
5541                if len(expressions) > 1:
5542                    return self.expression(exp.Concat, expressions=expressions)
5543
5544            return primary
5545
5546        if self._match_pair(TokenType.DOT, TokenType.NUMBER):
5547            return exp.Literal.number(f"0.{self._prev.text}")
5548
5549        if self._match(TokenType.L_PAREN):
5550            comments = self._prev_comments
5551            query = self._parse_select()
5552
5553            if query:
5554                expressions = [query]
5555            else:
5556                expressions = self._parse_expressions()
5557
5558            this = self._parse_query_modifiers(seq_get(expressions, 0))
5559
5560            if not this and self._match(TokenType.R_PAREN, advance=False):
5561                this = self.expression(exp.Tuple)
5562            elif isinstance(this, exp.UNWRAPPED_QUERIES):
5563                this = self._parse_subquery(this=this, parse_alias=False)
5564            elif isinstance(this, exp.Subquery):
5565                this = self._parse_subquery(
5566                    this=self._parse_set_operations(this), parse_alias=False
5567                )
5568            elif len(expressions) > 1 or self._prev.token_type == TokenType.COMMA:
5569                this = self.expression(exp.Tuple, expressions=expressions)
5570            else:
5571                this = self.expression(exp.Paren, this=this)
5572
5573            if this:
5574                this.add_comments(comments)
5575
5576            self._match_r_paren(expression=this)
5577            return this
5578
5579        return None
5580
5581    def _parse_field(
5582        self,
5583        any_token: bool = False,
5584        tokens: t.Optional[t.Collection[TokenType]] = None,
5585        anonymous_func: bool = False,
5586    ) -> t.Optional[exp.Expression]:
5587        if anonymous_func:
5588            field = (
5589                self._parse_function(anonymous=anonymous_func, any_token=any_token)
5590                or self._parse_primary()
5591            )
5592        else:
5593            field = self._parse_primary() or self._parse_function(
5594                anonymous=anonymous_func, any_token=any_token
5595            )
5596        return field or self._parse_id_var(any_token=any_token, tokens=tokens)
5597
5598    def _parse_function(
5599        self,
5600        functions: t.Optional[t.Dict[str, t.Callable]] = None,
5601        anonymous: bool = False,
5602        optional_parens: bool = True,
5603        any_token: bool = False,
5604    ) -> t.Optional[exp.Expression]:
5605        # This allows us to also parse {fn <function>} syntax (Snowflake, MySQL support this)
5606        # See: https://community.snowflake.com/s/article/SQL-Escape-Sequences
5607        fn_syntax = False
5608        if (
5609            self._match(TokenType.L_BRACE, advance=False)
5610            and self._next
5611            and self._next.text.upper() == "FN"
5612        ):
5613            self._advance(2)
5614            fn_syntax = True
5615
5616        func = self._parse_function_call(
5617            functions=functions,
5618            anonymous=anonymous,
5619            optional_parens=optional_parens,
5620            any_token=any_token,
5621        )
5622
5623        if fn_syntax:
5624            self._match(TokenType.R_BRACE)
5625
5626        return func
5627
5628    def _parse_function_call(
5629        self,
5630        functions: t.Optional[t.Dict[str, t.Callable]] = None,
5631        anonymous: bool = False,
5632        optional_parens: bool = True,
5633        any_token: bool = False,
5634    ) -> t.Optional[exp.Expression]:
5635        if not self._curr:
5636            return None
5637
5638        comments = self._curr.comments
5639        token = self._curr
5640        token_type = self._curr.token_type
5641        this = self._curr.text
5642        upper = this.upper()
5643
5644        parser = self.NO_PAREN_FUNCTION_PARSERS.get(upper)
5645        if optional_parens and parser and token_type not in self.INVALID_FUNC_NAME_TOKENS:
5646            self._advance()
5647            return self._parse_window(parser(self))
5648
5649        if not self._next or self._next.token_type != TokenType.L_PAREN:
5650            if optional_parens and token_type in self.NO_PAREN_FUNCTIONS:
5651                self._advance()
5652                return self.expression(self.NO_PAREN_FUNCTIONS[token_type])
5653
5654            return None
5655
5656        if any_token:
5657            if token_type in self.RESERVED_TOKENS:
5658                return None
5659        elif token_type not in self.FUNC_TOKENS:
5660            return None
5661
5662        self._advance(2)
5663
5664        parser = self.FUNCTION_PARSERS.get(upper)
5665        if parser and not anonymous:
5666            this = parser(self)
5667        else:
5668            subquery_predicate = self.SUBQUERY_PREDICATES.get(token_type)
5669
5670            if subquery_predicate and self._curr.token_type in (TokenType.SELECT, TokenType.WITH):
5671                this = self.expression(
5672                    subquery_predicate, comments=comments, this=self._parse_select()
5673                )
5674                self._match_r_paren()
5675                return this
5676
5677            if functions is None:
5678                functions = self.FUNCTIONS
5679
5680            function = functions.get(upper)
5681            known_function = function and not anonymous
5682
5683            alias = not known_function or upper in self.FUNCTIONS_WITH_ALIASED_ARGS
5684            args = self._parse_csv(lambda: self._parse_lambda(alias=alias))
5685
5686            post_func_comments = self._curr and self._curr.comments
5687            if known_function and post_func_comments:
5688                # If the user-inputted comment "/* sqlglot.anonymous */" is following the function
5689                # call we'll construct it as exp.Anonymous, even if it's "known"
5690                if any(
5691                    comment.lstrip().startswith(exp.SQLGLOT_ANONYMOUS)
5692                    for comment in post_func_comments
5693                ):
5694                    known_function = False
5695
5696            if alias and known_function:
5697                args = self._kv_to_prop_eq(args)
5698
5699            if known_function:
5700                func_builder = t.cast(t.Callable, function)
5701
5702                if "dialect" in func_builder.__code__.co_varnames:
5703                    func = func_builder(args, dialect=self.dialect)
5704                else:
5705                    func = func_builder(args)
5706
5707                func = self.validate_expression(func, args)
5708                if self.dialect.PRESERVE_ORIGINAL_NAMES:
5709                    func.meta["name"] = this
5710
5711                this = func
5712            else:
5713                if token_type == TokenType.IDENTIFIER:
5714                    this = exp.Identifier(this=this, quoted=True).update_positions(token)
5715                this = self.expression(exp.Anonymous, this=this, expressions=args)
5716
5717        if isinstance(this, exp.Expression):
5718            this.add_comments(comments)
5719
5720        self._match_r_paren(this)
5721        return self._parse_window(this)
5722
5723    def _to_prop_eq(self, expression: exp.Expression, index: int) -> exp.Expression:
5724        return expression
5725
5726    def _kv_to_prop_eq(self, expressions: t.List[exp.Expression]) -> t.List[exp.Expression]:
5727        transformed = []
5728
5729        for index, e in enumerate(expressions):
5730            if isinstance(e, self.KEY_VALUE_DEFINITIONS):
5731                if isinstance(e, exp.Alias):
5732                    e = self.expression(exp.PropertyEQ, this=e.args.get("alias"), expression=e.this)
5733
5734                if not isinstance(e, exp.PropertyEQ):
5735                    e = self.expression(
5736                        exp.PropertyEQ, this=exp.to_identifier(e.this.name), expression=e.expression
5737                    )
5738
5739                if isinstance(e.this, exp.Column):
5740                    e.this.replace(e.this.this)
5741            else:
5742                e = self._to_prop_eq(e, index)
5743
5744            transformed.append(e)
5745
5746        return transformed
5747
5748    def _parse_user_defined_function_expression(self) -> t.Optional[exp.Expression]:
5749        return self._parse_statement()
5750
5751    def _parse_function_parameter(self) -> t.Optional[exp.Expression]:
5752        return self._parse_column_def(this=self._parse_id_var(), computed_column=False)
5753
5754    def _parse_user_defined_function(
5755        self, kind: t.Optional[TokenType] = None
5756    ) -> t.Optional[exp.Expression]:
5757        this = self._parse_table_parts(schema=True)
5758
5759        if not self._match(TokenType.L_PAREN):
5760            return this
5761
5762        expressions = self._parse_csv(self._parse_function_parameter)
5763        self._match_r_paren()
5764        return self.expression(
5765            exp.UserDefinedFunction, this=this, expressions=expressions, wrapped=True
5766        )
5767
5768    def _parse_introducer(self, token: Token) -> exp.Introducer | exp.Identifier:
5769        literal = self._parse_primary()
5770        if literal:
5771            return self.expression(exp.Introducer, this=token.text, expression=literal)
5772
5773        return self._identifier_expression(token)
5774
5775    def _parse_session_parameter(self) -> exp.SessionParameter:
5776        kind = None
5777        this = self._parse_id_var() or self._parse_primary()
5778
5779        if this and self._match(TokenType.DOT):
5780            kind = this.name
5781            this = self._parse_var() or self._parse_primary()
5782
5783        return self.expression(exp.SessionParameter, this=this, kind=kind)
5784
5785    def _parse_lambda_arg(self) -> t.Optional[exp.Expression]:
5786        return self._parse_id_var()
5787
5788    def _parse_lambda(self, alias: bool = False) -> t.Optional[exp.Expression]:
5789        index = self._index
5790
5791        if self._match(TokenType.L_PAREN):
5792            expressions = t.cast(
5793                t.List[t.Optional[exp.Expression]], self._parse_csv(self._parse_lambda_arg)
5794            )
5795
5796            if not self._match(TokenType.R_PAREN):
5797                self._retreat(index)
5798        else:
5799            expressions = [self._parse_lambda_arg()]
5800
5801        if self._match_set(self.LAMBDAS):
5802            return self.LAMBDAS[self._prev.token_type](self, expressions)
5803
5804        self._retreat(index)
5805
5806        this: t.Optional[exp.Expression]
5807
5808        if self._match(TokenType.DISTINCT):
5809            this = self.expression(
5810                exp.Distinct, expressions=self._parse_csv(self._parse_assignment)
5811            )
5812        else:
5813            this = self._parse_select_or_expression(alias=alias)
5814
5815        return self._parse_limit(
5816            self._parse_order(self._parse_having_max(self._parse_respect_or_ignore_nulls(this)))
5817        )
5818
5819    def _parse_schema(self, this: t.Optional[exp.Expression] = None) -> t.Optional[exp.Expression]:
5820        index = self._index
5821        if not self._match(TokenType.L_PAREN):
5822            return this
5823
5824        # Disambiguate between schema and subquery/CTE, e.g. in INSERT INTO table (<expr>),
5825        # expr can be of both types
5826        if self._match_set(self.SELECT_START_TOKENS):
5827            self._retreat(index)
5828            return this
5829        args = self._parse_csv(lambda: self._parse_constraint() or self._parse_field_def())
5830        self._match_r_paren()
5831        return self.expression(exp.Schema, this=this, expressions=args)
5832
5833    def _parse_field_def(self) -> t.Optional[exp.Expression]:
5834        return self._parse_column_def(self._parse_field(any_token=True))
5835
5836    def _parse_column_def(
5837        self, this: t.Optional[exp.Expression], computed_column: bool = True
5838    ) -> t.Optional[exp.Expression]:
5839        # column defs are not really columns, they're identifiers
5840        if isinstance(this, exp.Column):
5841            this = this.this
5842
5843        if not computed_column:
5844            self._match(TokenType.ALIAS)
5845
5846        kind = self._parse_types(schema=True)
5847
5848        if self._match_text_seq("FOR", "ORDINALITY"):
5849            return self.expression(exp.ColumnDef, this=this, ordinality=True)
5850
5851        constraints: t.List[exp.Expression] = []
5852
5853        if (not kind and self._match(TokenType.ALIAS)) or self._match_texts(
5854            ("ALIAS", "MATERIALIZED")
5855        ):
5856            persisted = self._prev.text.upper() == "MATERIALIZED"
5857            constraint_kind = exp.ComputedColumnConstraint(
5858                this=self._parse_assignment(),
5859                persisted=persisted or self._match_text_seq("PERSISTED"),
5860                not_null=self._match_pair(TokenType.NOT, TokenType.NULL),
5861            )
5862            constraints.append(self.expression(exp.ColumnConstraint, kind=constraint_kind))
5863        elif (
5864            kind
5865            and self._match(TokenType.ALIAS, advance=False)
5866            and (
5867                not self.WRAPPED_TRANSFORM_COLUMN_CONSTRAINT
5868                or (self._next and self._next.token_type == TokenType.L_PAREN)
5869            )
5870        ):
5871            self._advance()
5872            constraints.append(
5873                self.expression(
5874                    exp.ColumnConstraint,
5875                    kind=exp.TransformColumnConstraint(this=self._parse_disjunction()),
5876                )
5877            )
5878
5879        while True:
5880            constraint = self._parse_column_constraint()
5881            if not constraint:
5882                break
5883            constraints.append(constraint)
5884
5885        if not kind and not constraints:
5886            return this
5887
5888        return self.expression(exp.ColumnDef, this=this, kind=kind, constraints=constraints)
5889
5890    def _parse_auto_increment(
5891        self,
5892    ) -> exp.GeneratedAsIdentityColumnConstraint | exp.AutoIncrementColumnConstraint:
5893        start = None
5894        increment = None
5895
5896        if self._match(TokenType.L_PAREN, advance=False):
5897            args = self._parse_wrapped_csv(self._parse_bitwise)
5898            start = seq_get(args, 0)
5899            increment = seq_get(args, 1)
5900        elif self._match_text_seq("START"):
5901            start = self._parse_bitwise()
5902            self._match_text_seq("INCREMENT")
5903            increment = self._parse_bitwise()
5904
5905        if start and increment:
5906            return exp.GeneratedAsIdentityColumnConstraint(
5907                start=start, increment=increment, this=False
5908            )
5909
5910        return exp.AutoIncrementColumnConstraint()
5911
5912    def _parse_auto_property(self) -> t.Optional[exp.AutoRefreshProperty]:
5913        if not self._match_text_seq("REFRESH"):
5914            self._retreat(self._index - 1)
5915            return None
5916        return self.expression(exp.AutoRefreshProperty, this=self._parse_var(upper=True))
5917
5918    def _parse_compress(self) -> exp.CompressColumnConstraint:
5919        if self._match(TokenType.L_PAREN, advance=False):
5920            return self.expression(
5921                exp.CompressColumnConstraint, this=self._parse_wrapped_csv(self._parse_bitwise)
5922            )
5923
5924        return self.expression(exp.CompressColumnConstraint, this=self._parse_bitwise())
5925
5926    def _parse_generated_as_identity(
5927        self,
5928    ) -> (
5929        exp.GeneratedAsIdentityColumnConstraint
5930        | exp.ComputedColumnConstraint
5931        | exp.GeneratedAsRowColumnConstraint
5932    ):
5933        if self._match_text_seq("BY", "DEFAULT"):
5934            on_null = self._match_pair(TokenType.ON, TokenType.NULL)
5935            this = self.expression(
5936                exp.GeneratedAsIdentityColumnConstraint, this=False, on_null=on_null
5937            )
5938        else:
5939            self._match_text_seq("ALWAYS")
5940            this = self.expression(exp.GeneratedAsIdentityColumnConstraint, this=True)
5941
5942        self._match(TokenType.ALIAS)
5943
5944        if self._match_text_seq("ROW"):
5945            start = self._match_text_seq("START")
5946            if not start:
5947                self._match(TokenType.END)
5948            hidden = self._match_text_seq("HIDDEN")
5949            return self.expression(exp.GeneratedAsRowColumnConstraint, start=start, hidden=hidden)
5950
5951        identity = self._match_text_seq("IDENTITY")
5952
5953        if self._match(TokenType.L_PAREN):
5954            if self._match(TokenType.START_WITH):
5955                this.set("start", self._parse_bitwise())
5956            if self._match_text_seq("INCREMENT", "BY"):
5957                this.set("increment", self._parse_bitwise())
5958            if self._match_text_seq("MINVALUE"):
5959                this.set("minvalue", self._parse_bitwise())
5960            if self._match_text_seq("MAXVALUE"):
5961                this.set("maxvalue", self._parse_bitwise())
5962
5963            if self._match_text_seq("CYCLE"):
5964                this.set("cycle", True)
5965            elif self._match_text_seq("NO", "CYCLE"):
5966                this.set("cycle", False)
5967
5968            if not identity:
5969                this.set("expression", self._parse_range())
5970            elif not this.args.get("start") and self._match(TokenType.NUMBER, advance=False):
5971                args = self._parse_csv(self._parse_bitwise)
5972                this.set("start", seq_get(args, 0))
5973                this.set("increment", seq_get(args, 1))
5974
5975            self._match_r_paren()
5976
5977        return this
5978
5979    def _parse_inline(self) -> exp.InlineLengthColumnConstraint:
5980        self._match_text_seq("LENGTH")
5981        return self.expression(exp.InlineLengthColumnConstraint, this=self._parse_bitwise())
5982
5983    def _parse_not_constraint(self) -> t.Optional[exp.Expression]:
5984        if self._match_text_seq("NULL"):
5985            return self.expression(exp.NotNullColumnConstraint)
5986        if self._match_text_seq("CASESPECIFIC"):
5987            return self.expression(exp.CaseSpecificColumnConstraint, not_=True)
5988        if self._match_text_seq("FOR", "REPLICATION"):
5989            return self.expression(exp.NotForReplicationColumnConstraint)
5990
5991        # Unconsume the `NOT` token
5992        self._retreat(self._index - 1)
5993        return None
5994
5995    def _parse_column_constraint(self) -> t.Optional[exp.Expression]:
5996        this = self._match(TokenType.CONSTRAINT) and self._parse_id_var()
5997
5998        procedure_option_follows = (
5999            self._match(TokenType.WITH, advance=False)
6000            and self._next
6001            and self._next.text.upper() in self.PROCEDURE_OPTIONS
6002        )
6003
6004        if not procedure_option_follows and self._match_texts(self.CONSTRAINT_PARSERS):
6005            return self.expression(
6006                exp.ColumnConstraint,
6007                this=this,
6008                kind=self.CONSTRAINT_PARSERS[self._prev.text.upper()](self),
6009            )
6010
6011        return this
6012
6013    def _parse_constraint(self) -> t.Optional[exp.Expression]:
6014        if not self._match(TokenType.CONSTRAINT):
6015            return self._parse_unnamed_constraint(constraints=self.SCHEMA_UNNAMED_CONSTRAINTS)
6016
6017        return self.expression(
6018            exp.Constraint,
6019            this=self._parse_id_var(),
6020            expressions=self._parse_unnamed_constraints(),
6021        )
6022
6023    def _parse_unnamed_constraints(self) -> t.List[exp.Expression]:
6024        constraints = []
6025        while True:
6026            constraint = self._parse_unnamed_constraint() or self._parse_function()
6027            if not constraint:
6028                break
6029            constraints.append(constraint)
6030
6031        return constraints
6032
6033    def _parse_unnamed_constraint(
6034        self, constraints: t.Optional[t.Collection[str]] = None
6035    ) -> t.Optional[exp.Expression]:
6036        if self._match(TokenType.IDENTIFIER, advance=False) or not self._match_texts(
6037            constraints or self.CONSTRAINT_PARSERS
6038        ):
6039            return None
6040
6041        constraint = self._prev.text.upper()
6042        if constraint not in self.CONSTRAINT_PARSERS:
6043            self.raise_error(f"No parser found for schema constraint {constraint}.")
6044
6045        return self.CONSTRAINT_PARSERS[constraint](self)
6046
6047    def _parse_unique_key(self) -> t.Optional[exp.Expression]:
6048        return self._parse_id_var(any_token=False)
6049
6050    def _parse_unique(self) -> exp.UniqueColumnConstraint:
6051        self._match_text_seq("KEY")
6052        return self.expression(
6053            exp.UniqueColumnConstraint,
6054            nulls=self._match_text_seq("NULLS", "NOT", "DISTINCT"),
6055            this=self._parse_schema(self._parse_unique_key()),
6056            index_type=self._match(TokenType.USING) and self._advance_any() and self._prev.text,
6057            on_conflict=self._parse_on_conflict(),
6058            options=self._parse_key_constraint_options(),
6059        )
6060
6061    def _parse_key_constraint_options(self) -> t.List[str]:
6062        options = []
6063        while True:
6064            if not self._curr:
6065                break
6066
6067            if self._match(TokenType.ON):
6068                action = None
6069                on = self._advance_any() and self._prev.text
6070
6071                if self._match_text_seq("NO", "ACTION"):
6072                    action = "NO ACTION"
6073                elif self._match_text_seq("CASCADE"):
6074                    action = "CASCADE"
6075                elif self._match_text_seq("RESTRICT"):
6076                    action = "RESTRICT"
6077                elif self._match_pair(TokenType.SET, TokenType.NULL):
6078                    action = "SET NULL"
6079                elif self._match_pair(TokenType.SET, TokenType.DEFAULT):
6080                    action = "SET DEFAULT"
6081                else:
6082                    self.raise_error("Invalid key constraint")
6083
6084                options.append(f"ON {on} {action}")
6085            else:
6086                var = self._parse_var_from_options(
6087                    self.KEY_CONSTRAINT_OPTIONS, raise_unmatched=False
6088                )
6089                if not var:
6090                    break
6091                options.append(var.name)
6092
6093        return options
6094
6095    def _parse_references(self, match: bool = True) -> t.Optional[exp.Reference]:
6096        if match and not self._match(TokenType.REFERENCES):
6097            return None
6098
6099        expressions = None
6100        this = self._parse_table(schema=True)
6101        options = self._parse_key_constraint_options()
6102        return self.expression(exp.Reference, this=this, expressions=expressions, options=options)
6103
6104    def _parse_foreign_key(self) -> exp.ForeignKey:
6105        expressions = (
6106            self._parse_wrapped_id_vars()
6107            if not self._match(TokenType.REFERENCES, advance=False)
6108            else None
6109        )
6110        reference = self._parse_references()
6111        on_options = {}
6112
6113        while self._match(TokenType.ON):
6114            if not self._match_set((TokenType.DELETE, TokenType.UPDATE)):
6115                self.raise_error("Expected DELETE or UPDATE")
6116
6117            kind = self._prev.text.lower()
6118
6119            if self._match_text_seq("NO", "ACTION"):
6120                action = "NO ACTION"
6121            elif self._match(TokenType.SET):
6122                self._match_set((TokenType.NULL, TokenType.DEFAULT))
6123                action = "SET " + self._prev.text.upper()
6124            else:
6125                self._advance()
6126                action = self._prev.text.upper()
6127
6128            on_options[kind] = action
6129
6130        return self.expression(
6131            exp.ForeignKey,
6132            expressions=expressions,
6133            reference=reference,
6134            options=self._parse_key_constraint_options(),
6135            **on_options,  # type: ignore
6136        )
6137
6138    def _parse_primary_key_part(self) -> t.Optional[exp.Expression]:
6139        return self._parse_ordered() or self._parse_field()
6140
6141    def _parse_period_for_system_time(self) -> t.Optional[exp.PeriodForSystemTimeConstraint]:
6142        if not self._match(TokenType.TIMESTAMP_SNAPSHOT):
6143            self._retreat(self._index - 1)
6144            return None
6145
6146        id_vars = self._parse_wrapped_id_vars()
6147        return self.expression(
6148            exp.PeriodForSystemTimeConstraint,
6149            this=seq_get(id_vars, 0),
6150            expression=seq_get(id_vars, 1),
6151        )
6152
6153    def _parse_primary_key(
6154        self, wrapped_optional: bool = False, in_props: bool = False
6155    ) -> exp.PrimaryKeyColumnConstraint | exp.PrimaryKey:
6156        desc = (
6157            self._match_set((TokenType.ASC, TokenType.DESC))
6158            and self._prev.token_type == TokenType.DESC
6159        )
6160
6161        if not in_props and not self._match(TokenType.L_PAREN, advance=False):
6162            return self.expression(
6163                exp.PrimaryKeyColumnConstraint,
6164                desc=desc,
6165                options=self._parse_key_constraint_options(),
6166            )
6167
6168        expressions = self._parse_wrapped_csv(
6169            self._parse_primary_key_part, optional=wrapped_optional
6170        )
6171        options = self._parse_key_constraint_options()
6172        return self.expression(exp.PrimaryKey, expressions=expressions, options=options)
6173
6174    def _parse_bracket_key_value(self, is_map: bool = False) -> t.Optional[exp.Expression]:
6175        return self._parse_slice(self._parse_alias(self._parse_assignment(), explicit=True))
6176
6177    def _parse_odbc_datetime_literal(self) -> exp.Expression:
6178        """
6179        Parses a datetime column in ODBC format. We parse the column into the corresponding
6180        types, for example `{d'yyyy-mm-dd'}` will be parsed as a `Date` column, exactly the
6181        same as we did for `DATE('yyyy-mm-dd')`.
6182
6183        Reference:
6184        https://learn.microsoft.com/en-us/sql/odbc/reference/develop-app/date-time-and-timestamp-literals
6185        """
6186        self._match(TokenType.VAR)
6187        exp_class = self.ODBC_DATETIME_LITERALS[self._prev.text.lower()]
6188        expression = self.expression(exp_class=exp_class, this=self._parse_string())
6189        if not self._match(TokenType.R_BRACE):
6190            self.raise_error("Expected }")
6191        return expression
6192
6193    def _parse_bracket(self, this: t.Optional[exp.Expression] = None) -> t.Optional[exp.Expression]:
6194        if not self._match_set((TokenType.L_BRACKET, TokenType.L_BRACE)):
6195            return this
6196
6197        bracket_kind = self._prev.token_type
6198        if (
6199            bracket_kind == TokenType.L_BRACE
6200            and self._curr
6201            and self._curr.token_type == TokenType.VAR
6202            and self._curr.text.lower() in self.ODBC_DATETIME_LITERALS
6203        ):
6204            return self._parse_odbc_datetime_literal()
6205
6206        expressions = self._parse_csv(
6207            lambda: self._parse_bracket_key_value(is_map=bracket_kind == TokenType.L_BRACE)
6208        )
6209
6210        if bracket_kind == TokenType.L_BRACKET and not self._match(TokenType.R_BRACKET):
6211            self.raise_error("Expected ]")
6212        elif bracket_kind == TokenType.L_BRACE and not self._match(TokenType.R_BRACE):
6213            self.raise_error("Expected }")
6214
6215        # https://duckdb.org/docs/sql/data_types/struct.html#creating-structs
6216        if bracket_kind == TokenType.L_BRACE:
6217            this = self.expression(exp.Struct, expressions=self._kv_to_prop_eq(expressions))
6218        elif not this:
6219            this = build_array_constructor(
6220                exp.Array, args=expressions, bracket_kind=bracket_kind, dialect=self.dialect
6221            )
6222        else:
6223            constructor_type = self.ARRAY_CONSTRUCTORS.get(this.name.upper())
6224            if constructor_type:
6225                return build_array_constructor(
6226                    constructor_type,
6227                    args=expressions,
6228                    bracket_kind=bracket_kind,
6229                    dialect=self.dialect,
6230                )
6231
6232            expressions = apply_index_offset(
6233                this, expressions, -self.dialect.INDEX_OFFSET, dialect=self.dialect
6234            )
6235            this = self.expression(exp.Bracket, this=this, expressions=expressions)
6236
6237        self._add_comments(this)
6238        return self._parse_bracket(this)
6239
6240    def _parse_slice(self, this: t.Optional[exp.Expression]) -> t.Optional[exp.Expression]:
6241        if self._match(TokenType.COLON):
6242            return self.expression(exp.Slice, this=this, expression=self._parse_assignment())
6243        return this
6244
6245    def _parse_case(self) -> t.Optional[exp.Expression]:
6246        ifs = []
6247        default = None
6248
6249        comments = self._prev_comments
6250        expression = self._parse_assignment()
6251
6252        while self._match(TokenType.WHEN):
6253            this = self._parse_assignment()
6254            self._match(TokenType.THEN)
6255            then = self._parse_assignment()
6256            ifs.append(self.expression(exp.If, this=this, true=then))
6257
6258        if self._match(TokenType.ELSE):
6259            default = self._parse_assignment()
6260
6261        if not self._match(TokenType.END):
6262            if isinstance(default, exp.Interval) and default.this.sql().upper() == "END":
6263                default = exp.column("interval")
6264            else:
6265                self.raise_error("Expected END after CASE", self._prev)
6266
6267        return self.expression(
6268            exp.Case, comments=comments, this=expression, ifs=ifs, default=default
6269        )
6270
6271    def _parse_if(self) -> t.Optional[exp.Expression]:
6272        if self._match(TokenType.L_PAREN):
6273            args = self._parse_csv(
6274                lambda: self._parse_alias(self._parse_assignment(), explicit=True)
6275            )
6276            this = self.validate_expression(exp.If.from_arg_list(args), args)
6277            self._match_r_paren()
6278        else:
6279            index = self._index - 1
6280
6281            if self.NO_PAREN_IF_COMMANDS and index == 0:
6282                return self._parse_as_command(self._prev)
6283
6284            condition = self._parse_assignment()
6285
6286            if not condition:
6287                self._retreat(index)
6288                return None
6289
6290            self._match(TokenType.THEN)
6291            true = self._parse_assignment()
6292            false = self._parse_assignment() if self._match(TokenType.ELSE) else None
6293            self._match(TokenType.END)
6294            this = self.expression(exp.If, this=condition, true=true, false=false)
6295
6296        return this
6297
6298    def _parse_next_value_for(self) -> t.Optional[exp.Expression]:
6299        if not self._match_text_seq("VALUE", "FOR"):
6300            self._retreat(self._index - 1)
6301            return None
6302
6303        return self.expression(
6304            exp.NextValueFor,
6305            this=self._parse_column(),
6306            order=self._match(TokenType.OVER) and self._parse_wrapped(self._parse_order),
6307        )
6308
6309    def _parse_extract(self) -> exp.Extract | exp.Anonymous:
6310        this = self._parse_function() or self._parse_var_or_string(upper=True)
6311
6312        if self._match(TokenType.FROM):
6313            return self.expression(exp.Extract, this=this, expression=self._parse_bitwise())
6314
6315        if not self._match(TokenType.COMMA):
6316            self.raise_error("Expected FROM or comma after EXTRACT", self._prev)
6317
6318        return self.expression(exp.Extract, this=this, expression=self._parse_bitwise())
6319
6320    def _parse_gap_fill(self) -> exp.GapFill:
6321        self._match(TokenType.TABLE)
6322        this = self._parse_table()
6323
6324        self._match(TokenType.COMMA)
6325        args = [this, *self._parse_csv(self._parse_lambda)]
6326
6327        gap_fill = exp.GapFill.from_arg_list(args)
6328        return self.validate_expression(gap_fill, args)
6329
6330    def _parse_cast(self, strict: bool, safe: t.Optional[bool] = None) -> exp.Expression:
6331        this = self._parse_assignment()
6332
6333        if not self._match(TokenType.ALIAS):
6334            if self._match(TokenType.COMMA):
6335                return self.expression(exp.CastToStrType, this=this, to=self._parse_string())
6336
6337            self.raise_error("Expected AS after CAST")
6338
6339        fmt = None
6340        to = self._parse_types()
6341
6342        default = self._match(TokenType.DEFAULT)
6343        if default:
6344            default = self._parse_bitwise()
6345            self._match_text_seq("ON", "CONVERSION", "ERROR")
6346
6347        if self._match_set((TokenType.FORMAT, TokenType.COMMA)):
6348            fmt_string = self._parse_string()
6349            fmt = self._parse_at_time_zone(fmt_string)
6350
6351            if not to:
6352                to = exp.DataType.build(exp.DataType.Type.UNKNOWN)
6353            if to.this in exp.DataType.TEMPORAL_TYPES:
6354                this = self.expression(
6355                    exp.StrToDate if to.this == exp.DataType.Type.DATE else exp.StrToTime,
6356                    this=this,
6357                    format=exp.Literal.string(
6358                        format_time(
6359                            fmt_string.this if fmt_string else "",
6360                            self.dialect.FORMAT_MAPPING or self.dialect.TIME_MAPPING,
6361                            self.dialect.FORMAT_TRIE or self.dialect.TIME_TRIE,
6362                        )
6363                    ),
6364                    safe=safe,
6365                )
6366
6367                if isinstance(fmt, exp.AtTimeZone) and isinstance(this, exp.StrToTime):
6368                    this.set("zone", fmt.args["zone"])
6369                return this
6370        elif not to:
6371            self.raise_error("Expected TYPE after CAST")
6372        elif isinstance(to, exp.Identifier):
6373            to = exp.DataType.build(to.name, udt=True)
6374        elif to.this == exp.DataType.Type.CHAR:
6375            if self._match(TokenType.CHARACTER_SET):
6376                to = self.expression(exp.CharacterSet, this=self._parse_var_or_string())
6377
6378        return self.expression(
6379            exp.Cast if strict else exp.TryCast,
6380            this=this,
6381            to=to,
6382            format=fmt,
6383            safe=safe,
6384            action=self._parse_var_from_options(self.CAST_ACTIONS, raise_unmatched=False),
6385            default=default,
6386        )
6387
6388    def _parse_string_agg(self) -> exp.GroupConcat:
6389        if self._match(TokenType.DISTINCT):
6390            args: t.List[t.Optional[exp.Expression]] = [
6391                self.expression(exp.Distinct, expressions=[self._parse_assignment()])
6392            ]
6393            if self._match(TokenType.COMMA):
6394                args.extend(self._parse_csv(self._parse_assignment))
6395        else:
6396            args = self._parse_csv(self._parse_assignment)  # type: ignore
6397
6398        if self._match_text_seq("ON", "OVERFLOW"):
6399            # trino: LISTAGG(expression [, separator] [ON OVERFLOW overflow_behavior])
6400            if self._match_text_seq("ERROR"):
6401                on_overflow: t.Optional[exp.Expression] = exp.var("ERROR")
6402            else:
6403                self._match_text_seq("TRUNCATE")
6404                on_overflow = self.expression(
6405                    exp.OverflowTruncateBehavior,
6406                    this=self._parse_string(),
6407                    with_count=(
6408                        self._match_text_seq("WITH", "COUNT")
6409                        or not self._match_text_seq("WITHOUT", "COUNT")
6410                    ),
6411                )
6412        else:
6413            on_overflow = None
6414
6415        index = self._index
6416        if not self._match(TokenType.R_PAREN) and args:
6417            # postgres: STRING_AGG([DISTINCT] expression, separator [ORDER BY expression1 {ASC | DESC} [, ...]])
6418            # bigquery: STRING_AGG([DISTINCT] expression [, separator] [ORDER BY key [{ASC | DESC}] [, ... ]] [LIMIT n])
6419            # The order is parsed through `this` as a canonicalization for WITHIN GROUPs
6420            args[0] = self._parse_limit(this=self._parse_order(this=args[0]))
6421            return self.expression(exp.GroupConcat, this=args[0], separator=seq_get(args, 1))
6422
6423        # Checks if we can parse an order clause: WITHIN GROUP (ORDER BY <order_by_expression_list> [ASC | DESC]).
6424        # This is done "manually", instead of letting _parse_window parse it into an exp.WithinGroup node, so that
6425        # the STRING_AGG call is parsed like in MySQL / SQLite and can thus be transpiled more easily to them.
6426        if not self._match_text_seq("WITHIN", "GROUP"):
6427            self._retreat(index)
6428            return self.validate_expression(exp.GroupConcat.from_arg_list(args), args)
6429
6430        # The corresponding match_r_paren will be called in parse_function (caller)
6431        self._match_l_paren()
6432
6433        return self.expression(
6434            exp.GroupConcat,
6435            this=self._parse_order(this=seq_get(args, 0)),
6436            separator=seq_get(args, 1),
6437            on_overflow=on_overflow,
6438        )
6439
6440    def _parse_convert(
6441        self, strict: bool, safe: t.Optional[bool] = None
6442    ) -> t.Optional[exp.Expression]:
6443        this = self._parse_bitwise()
6444
6445        if self._match(TokenType.USING):
6446            to: t.Optional[exp.Expression] = self.expression(
6447                exp.CharacterSet, this=self._parse_var()
6448            )
6449        elif self._match(TokenType.COMMA):
6450            to = self._parse_types()
6451        else:
6452            to = None
6453
6454        return self.expression(exp.Cast if strict else exp.TryCast, this=this, to=to, safe=safe)
6455
6456    def _parse_xml_table(self) -> exp.XMLTable:
6457        namespaces = None
6458        passing = None
6459        columns = None
6460
6461        if self._match_text_seq("XMLNAMESPACES", "("):
6462            namespaces = self._parse_xml_namespace()
6463            self._match_text_seq(")", ",")
6464
6465        this = self._parse_string()
6466
6467        if self._match_text_seq("PASSING"):
6468            # The BY VALUE keywords are optional and are provided for semantic clarity
6469            self._match_text_seq("BY", "VALUE")
6470            passing = self._parse_csv(self._parse_column)
6471
6472        by_ref = self._match_text_seq("RETURNING", "SEQUENCE", "BY", "REF")
6473
6474        if self._match_text_seq("COLUMNS"):
6475            columns = self._parse_csv(self._parse_field_def)
6476
6477        return self.expression(
6478            exp.XMLTable,
6479            this=this,
6480            namespaces=namespaces,
6481            passing=passing,
6482            columns=columns,
6483            by_ref=by_ref,
6484        )
6485
6486    def _parse_xml_namespace(self) -> t.List[exp.XMLNamespace]:
6487        namespaces = []
6488
6489        while True:
6490            if self._match(TokenType.DEFAULT):
6491                uri = self._parse_string()
6492            else:
6493                uri = self._parse_alias(self._parse_string())
6494            namespaces.append(self.expression(exp.XMLNamespace, this=uri))
6495            if not self._match(TokenType.COMMA):
6496                break
6497
6498        return namespaces
6499
6500    def _parse_decode(self) -> t.Optional[exp.Decode | exp.Case]:
6501        """
6502        There are generally two variants of the DECODE function:
6503
6504        - DECODE(bin, charset)
6505        - DECODE(expression, search, result [, search, result] ... [, default])
6506
6507        The second variant will always be parsed into a CASE expression. Note that NULL
6508        needs special treatment, since we need to explicitly check for it with `IS NULL`,
6509        instead of relying on pattern matching.
6510        """
6511        args = self._parse_csv(self._parse_assignment)
6512
6513        if len(args) < 3:
6514            return self.expression(exp.Decode, this=seq_get(args, 0), charset=seq_get(args, 1))
6515
6516        expression, *expressions = args
6517        if not expression:
6518            return None
6519
6520        ifs = []
6521        for search, result in zip(expressions[::2], expressions[1::2]):
6522            if not search or not result:
6523                return None
6524
6525            if isinstance(search, exp.Literal):
6526                ifs.append(
6527                    exp.If(this=exp.EQ(this=expression.copy(), expression=search), true=result)
6528                )
6529            elif isinstance(search, exp.Null):
6530                ifs.append(
6531                    exp.If(this=exp.Is(this=expression.copy(), expression=exp.Null()), true=result)
6532                )
6533            else:
6534                cond = exp.or_(
6535                    exp.EQ(this=expression.copy(), expression=search),
6536                    exp.and_(
6537                        exp.Is(this=expression.copy(), expression=exp.Null()),
6538                        exp.Is(this=search.copy(), expression=exp.Null()),
6539                        copy=False,
6540                    ),
6541                    copy=False,
6542                )
6543                ifs.append(exp.If(this=cond, true=result))
6544
6545        return exp.Case(ifs=ifs, default=expressions[-1] if len(expressions) % 2 == 1 else None)
6546
6547    def _parse_json_key_value(self) -> t.Optional[exp.JSONKeyValue]:
6548        self._match_text_seq("KEY")
6549        key = self._parse_column()
6550        self._match_set(self.JSON_KEY_VALUE_SEPARATOR_TOKENS)
6551        self._match_text_seq("VALUE")
6552        value = self._parse_bitwise()
6553
6554        if not key and not value:
6555            return None
6556        return self.expression(exp.JSONKeyValue, this=key, expression=value)
6557
6558    def _parse_format_json(self, this: t.Optional[exp.Expression]) -> t.Optional[exp.Expression]:
6559        if not this or not self._match_text_seq("FORMAT", "JSON"):
6560            return this
6561
6562        return self.expression(exp.FormatJson, this=this)
6563
6564    def _parse_on_condition(self) -> t.Optional[exp.OnCondition]:
6565        # MySQL uses "X ON EMPTY Y ON ERROR" (e.g. JSON_VALUE) while Oracle uses the opposite (e.g. JSON_EXISTS)
6566        if self.dialect.ON_CONDITION_EMPTY_BEFORE_ERROR:
6567            empty = self._parse_on_handling("EMPTY", *self.ON_CONDITION_TOKENS)
6568            error = self._parse_on_handling("ERROR", *self.ON_CONDITION_TOKENS)
6569        else:
6570            error = self._parse_on_handling("ERROR", *self.ON_CONDITION_TOKENS)
6571            empty = self._parse_on_handling("EMPTY", *self.ON_CONDITION_TOKENS)
6572
6573        null = self._parse_on_handling("NULL", *self.ON_CONDITION_TOKENS)
6574
6575        if not empty and not error and not null:
6576            return None
6577
6578        return self.expression(
6579            exp.OnCondition,
6580            empty=empty,
6581            error=error,
6582            null=null,
6583        )
6584
6585    def _parse_on_handling(
6586        self, on: str, *values: str
6587    ) -> t.Optional[str] | t.Optional[exp.Expression]:
6588        # Parses the "X ON Y" or "DEFAULT <expr> ON Y syntax, e.g. NULL ON NULL (Oracle, T-SQL, MySQL)
6589        for value in values:
6590            if self._match_text_seq(value, "ON", on):
6591                return f"{value} ON {on}"
6592
6593        index = self._index
6594        if self._match(TokenType.DEFAULT):
6595            default_value = self._parse_bitwise()
6596            if self._match_text_seq("ON", on):
6597                return default_value
6598
6599            self._retreat(index)
6600
6601        return None
6602
6603    @t.overload
6604    def _parse_json_object(self, agg: Lit[False]) -> exp.JSONObject: ...
6605
6606    @t.overload
6607    def _parse_json_object(self, agg: Lit[True]) -> exp.JSONObjectAgg: ...
6608
6609    def _parse_json_object(self, agg=False):
6610        star = self._parse_star()
6611        expressions = (
6612            [star]
6613            if star
6614            else self._parse_csv(lambda: self._parse_format_json(self._parse_json_key_value()))
6615        )
6616        null_handling = self._parse_on_handling("NULL", "NULL", "ABSENT")
6617
6618        unique_keys = None
6619        if self._match_text_seq("WITH", "UNIQUE"):
6620            unique_keys = True
6621        elif self._match_text_seq("WITHOUT", "UNIQUE"):
6622            unique_keys = False
6623
6624        self._match_text_seq("KEYS")
6625
6626        return_type = self._match_text_seq("RETURNING") and self._parse_format_json(
6627            self._parse_type()
6628        )
6629        encoding = self._match_text_seq("ENCODING") and self._parse_var()
6630
6631        return self.expression(
6632            exp.JSONObjectAgg if agg else exp.JSONObject,
6633            expressions=expressions,
6634            null_handling=null_handling,
6635            unique_keys=unique_keys,
6636            return_type=return_type,
6637            encoding=encoding,
6638        )
6639
6640    # Note: this is currently incomplete; it only implements the "JSON_value_column" part
6641    def _parse_json_column_def(self) -> exp.JSONColumnDef:
6642        if not self._match_text_seq("NESTED"):
6643            this = self._parse_id_var()
6644            kind = self._parse_types(allow_identifiers=False)
6645            nested = None
6646        else:
6647            this = None
6648            kind = None
6649            nested = True
6650
6651        path = self._match_text_seq("PATH") and self._parse_string()
6652        nested_schema = nested and self._parse_json_schema()
6653
6654        return self.expression(
6655            exp.JSONColumnDef,
6656            this=this,
6657            kind=kind,
6658            path=path,
6659            nested_schema=nested_schema,
6660        )
6661
6662    def _parse_json_schema(self) -> exp.JSONSchema:
6663        self._match_text_seq("COLUMNS")
6664        return self.expression(
6665            exp.JSONSchema,
6666            expressions=self._parse_wrapped_csv(self._parse_json_column_def, optional=True),
6667        )
6668
6669    def _parse_json_table(self) -> exp.JSONTable:
6670        this = self._parse_format_json(self._parse_bitwise())
6671        path = self._match(TokenType.COMMA) and self._parse_string()
6672        error_handling = self._parse_on_handling("ERROR", "ERROR", "NULL")
6673        empty_handling = self._parse_on_handling("EMPTY", "ERROR", "NULL")
6674        schema = self._parse_json_schema()
6675
6676        return exp.JSONTable(
6677            this=this,
6678            schema=schema,
6679            path=path,
6680            error_handling=error_handling,
6681            empty_handling=empty_handling,
6682        )
6683
6684    def _parse_match_against(self) -> exp.MatchAgainst:
6685        expressions = self._parse_csv(self._parse_column)
6686
6687        self._match_text_seq(")", "AGAINST", "(")
6688
6689        this = self._parse_string()
6690
6691        if self._match_text_seq("IN", "NATURAL", "LANGUAGE", "MODE"):
6692            modifier = "IN NATURAL LANGUAGE MODE"
6693            if self._match_text_seq("WITH", "QUERY", "EXPANSION"):
6694                modifier = f"{modifier} WITH QUERY EXPANSION"
6695        elif self._match_text_seq("IN", "BOOLEAN", "MODE"):
6696            modifier = "IN BOOLEAN MODE"
6697        elif self._match_text_seq("WITH", "QUERY", "EXPANSION"):
6698            modifier = "WITH QUERY EXPANSION"
6699        else:
6700            modifier = None
6701
6702        return self.expression(
6703            exp.MatchAgainst, this=this, expressions=expressions, modifier=modifier
6704        )
6705
6706    # https://learn.microsoft.com/en-us/sql/t-sql/functions/openjson-transact-sql?view=sql-server-ver16
6707    def _parse_open_json(self) -> exp.OpenJSON:
6708        this = self._parse_bitwise()
6709        path = self._match(TokenType.COMMA) and self._parse_string()
6710
6711        def _parse_open_json_column_def() -> exp.OpenJSONColumnDef:
6712            this = self._parse_field(any_token=True)
6713            kind = self._parse_types()
6714            path = self._parse_string()
6715            as_json = self._match_pair(TokenType.ALIAS, TokenType.JSON)
6716
6717            return self.expression(
6718                exp.OpenJSONColumnDef, this=this, kind=kind, path=path, as_json=as_json
6719            )
6720
6721        expressions = None
6722        if self._match_pair(TokenType.R_PAREN, TokenType.WITH):
6723            self._match_l_paren()
6724            expressions = self._parse_csv(_parse_open_json_column_def)
6725
6726        return self.expression(exp.OpenJSON, this=this, path=path, expressions=expressions)
6727
6728    def _parse_position(self, haystack_first: bool = False) -> exp.StrPosition:
6729        args = self._parse_csv(self._parse_bitwise)
6730
6731        if self._match(TokenType.IN):
6732            return self.expression(
6733                exp.StrPosition, this=self._parse_bitwise(), substr=seq_get(args, 0)
6734            )
6735
6736        if haystack_first:
6737            haystack = seq_get(args, 0)
6738            needle = seq_get(args, 1)
6739        else:
6740            haystack = seq_get(args, 1)
6741            needle = seq_get(args, 0)
6742
6743        return self.expression(
6744            exp.StrPosition, this=haystack, substr=needle, position=seq_get(args, 2)
6745        )
6746
6747    def _parse_predict(self) -> exp.Predict:
6748        self._match_text_seq("MODEL")
6749        this = self._parse_table()
6750
6751        self._match(TokenType.COMMA)
6752        self._match_text_seq("TABLE")
6753
6754        return self.expression(
6755            exp.Predict,
6756            this=this,
6757            expression=self._parse_table(),
6758            params_struct=self._match(TokenType.COMMA) and self._parse_bitwise(),
6759        )
6760
6761    def _parse_join_hint(self, func_name: str) -> exp.JoinHint:
6762        args = self._parse_csv(self._parse_table)
6763        return exp.JoinHint(this=func_name.upper(), expressions=args)
6764
6765    def _parse_substring(self) -> exp.Substring:
6766        # Postgres supports the form: substring(string [from int] [for int])
6767        # https://www.postgresql.org/docs/9.1/functions-string.html @ Table 9-6
6768
6769        args = t.cast(t.List[t.Optional[exp.Expression]], self._parse_csv(self._parse_bitwise))
6770
6771        if self._match(TokenType.FROM):
6772            args.append(self._parse_bitwise())
6773        if self._match(TokenType.FOR):
6774            if len(args) == 1:
6775                args.append(exp.Literal.number(1))
6776            args.append(self._parse_bitwise())
6777
6778        return self.validate_expression(exp.Substring.from_arg_list(args), args)
6779
6780    def _parse_trim(self) -> exp.Trim:
6781        # https://www.w3resource.com/sql/character-functions/trim.php
6782        # https://docs.oracle.com/javadb/10.8.3.0/ref/rreftrimfunc.html
6783
6784        position = None
6785        collation = None
6786        expression = None
6787
6788        if self._match_texts(self.TRIM_TYPES):
6789            position = self._prev.text.upper()
6790
6791        this = self._parse_bitwise()
6792        if self._match_set((TokenType.FROM, TokenType.COMMA)):
6793            invert_order = self._prev.token_type == TokenType.FROM or self.TRIM_PATTERN_FIRST
6794            expression = self._parse_bitwise()
6795
6796            if invert_order:
6797                this, expression = expression, this
6798
6799        if self._match(TokenType.COLLATE):
6800            collation = self._parse_bitwise()
6801
6802        return self.expression(
6803            exp.Trim, this=this, position=position, expression=expression, collation=collation
6804        )
6805
6806    def _parse_window_clause(self) -> t.Optional[t.List[exp.Expression]]:
6807        return self._match(TokenType.WINDOW) and self._parse_csv(self._parse_named_window)
6808
6809    def _parse_named_window(self) -> t.Optional[exp.Expression]:
6810        return self._parse_window(self._parse_id_var(), alias=True)
6811
6812    def _parse_respect_or_ignore_nulls(
6813        self, this: t.Optional[exp.Expression]
6814    ) -> t.Optional[exp.Expression]:
6815        if self._match_text_seq("IGNORE", "NULLS"):
6816            return self.expression(exp.IgnoreNulls, this=this)
6817        if self._match_text_seq("RESPECT", "NULLS"):
6818            return self.expression(exp.RespectNulls, this=this)
6819        return this
6820
6821    def _parse_having_max(self, this: t.Optional[exp.Expression]) -> t.Optional[exp.Expression]:
6822        if self._match(TokenType.HAVING):
6823            self._match_texts(("MAX", "MIN"))
6824            max = self._prev.text.upper() != "MIN"
6825            return self.expression(
6826                exp.HavingMax, this=this, expression=self._parse_column(), max=max
6827            )
6828
6829        return this
6830
6831    def _parse_window(
6832        self, this: t.Optional[exp.Expression], alias: bool = False
6833    ) -> t.Optional[exp.Expression]:
6834        func = this
6835        comments = func.comments if isinstance(func, exp.Expression) else None
6836
6837        # T-SQL allows the OVER (...) syntax after WITHIN GROUP.
6838        # https://learn.microsoft.com/en-us/sql/t-sql/functions/percentile-disc-transact-sql?view=sql-server-ver16
6839        if self._match_text_seq("WITHIN", "GROUP"):
6840            order = self._parse_wrapped(self._parse_order)
6841            this = self.expression(exp.WithinGroup, this=this, expression=order)
6842
6843        if self._match_pair(TokenType.FILTER, TokenType.L_PAREN):
6844            self._match(TokenType.WHERE)
6845            this = self.expression(
6846                exp.Filter, this=this, expression=self._parse_where(skip_where_token=True)
6847            )
6848            self._match_r_paren()
6849
6850        # SQL spec defines an optional [ { IGNORE | RESPECT } NULLS ] OVER
6851        # Some dialects choose to implement and some do not.
6852        # https://dev.mysql.com/doc/refman/8.0/en/window-function-descriptions.html
6853
6854        # There is some code above in _parse_lambda that handles
6855        #   SELECT FIRST_VALUE(TABLE.COLUMN IGNORE|RESPECT NULLS) OVER ...
6856
6857        # The below changes handle
6858        #   SELECT FIRST_VALUE(TABLE.COLUMN) IGNORE|RESPECT NULLS OVER ...
6859
6860        # Oracle allows both formats
6861        #   (https://docs.oracle.com/en/database/oracle/oracle-database/19/sqlrf/img_text/first_value.html)
6862        #   and Snowflake chose to do the same for familiarity
6863        #   https://docs.snowflake.com/en/sql-reference/functions/first_value.html#usage-notes
6864        if isinstance(this, exp.AggFunc):
6865            ignore_respect = this.find(exp.IgnoreNulls, exp.RespectNulls)
6866
6867            if ignore_respect and ignore_respect is not this:
6868                ignore_respect.replace(ignore_respect.this)
6869                this = self.expression(ignore_respect.__class__, this=this)
6870
6871        this = self._parse_respect_or_ignore_nulls(this)
6872
6873        # bigquery select from window x AS (partition by ...)
6874        if alias:
6875            over = None
6876            self._match(TokenType.ALIAS)
6877        elif not self._match_set(self.WINDOW_BEFORE_PAREN_TOKENS):
6878            return this
6879        else:
6880            over = self._prev.text.upper()
6881
6882        if comments and isinstance(func, exp.Expression):
6883            func.pop_comments()
6884
6885        if not self._match(TokenType.L_PAREN):
6886            return self.expression(
6887                exp.Window,
6888                comments=comments,
6889                this=this,
6890                alias=self._parse_id_var(False),
6891                over=over,
6892            )
6893
6894        window_alias = self._parse_id_var(any_token=False, tokens=self.WINDOW_ALIAS_TOKENS)
6895
6896        first = self._match(TokenType.FIRST)
6897        if self._match_text_seq("LAST"):
6898            first = False
6899
6900        partition, order = self._parse_partition_and_order()
6901        kind = self._match_set((TokenType.ROWS, TokenType.RANGE)) and self._prev.text
6902
6903        if kind:
6904            self._match(TokenType.BETWEEN)
6905            start = self._parse_window_spec()
6906            self._match(TokenType.AND)
6907            end = self._parse_window_spec()
6908            exclude = (
6909                self._parse_var_from_options(self.WINDOW_EXCLUDE_OPTIONS)
6910                if self._match_text_seq("EXCLUDE")
6911                else None
6912            )
6913
6914            spec = self.expression(
6915                exp.WindowSpec,
6916                kind=kind,
6917                start=start["value"],
6918                start_side=start["side"],
6919                end=end["value"],
6920                end_side=end["side"],
6921                exclude=exclude,
6922            )
6923        else:
6924            spec = None
6925
6926        self._match_r_paren()
6927
6928        window = self.expression(
6929            exp.Window,
6930            comments=comments,
6931            this=this,
6932            partition_by=partition,
6933            order=order,
6934            spec=spec,
6935            alias=window_alias,
6936            over=over,
6937            first=first,
6938        )
6939
6940        # This covers Oracle's FIRST/LAST syntax: aggregate KEEP (...) OVER (...)
6941        if self._match_set(self.WINDOW_BEFORE_PAREN_TOKENS, advance=False):
6942            return self._parse_window(window, alias=alias)
6943
6944        return window
6945
6946    def _parse_partition_and_order(
6947        self,
6948    ) -> t.Tuple[t.List[exp.Expression], t.Optional[exp.Expression]]:
6949        return self._parse_partition_by(), self._parse_order()
6950
6951    def _parse_window_spec(self) -> t.Dict[str, t.Optional[str | exp.Expression]]:
6952        self._match(TokenType.BETWEEN)
6953
6954        return {
6955            "value": (
6956                (self._match_text_seq("UNBOUNDED") and "UNBOUNDED")
6957                or (self._match_text_seq("CURRENT", "ROW") and "CURRENT ROW")
6958                or self._parse_bitwise()
6959            ),
6960            "side": self._match_texts(self.WINDOW_SIDES) and self._prev.text,
6961        }
6962
6963    def _parse_alias(
6964        self, this: t.Optional[exp.Expression], explicit: bool = False
6965    ) -> t.Optional[exp.Expression]:
6966        # In some dialects, LIMIT and OFFSET can act as both identifiers and keywords (clauses)
6967        # so this section tries to parse the clause version and if it fails, it treats the token
6968        # as an identifier (alias)
6969        if self._can_parse_limit_or_offset():
6970            return this
6971
6972        any_token = self._match(TokenType.ALIAS)
6973        comments = self._prev_comments or []
6974
6975        if explicit and not any_token:
6976            return this
6977
6978        if self._match(TokenType.L_PAREN):
6979            aliases = self.expression(
6980                exp.Aliases,
6981                comments=comments,
6982                this=this,
6983                expressions=self._parse_csv(lambda: self._parse_id_var(any_token)),
6984            )
6985            self._match_r_paren(aliases)
6986            return aliases
6987
6988        alias = self._parse_id_var(any_token, tokens=self.ALIAS_TOKENS) or (
6989            self.STRING_ALIASES and self._parse_string_as_identifier()
6990        )
6991
6992        if alias:
6993            comments.extend(alias.pop_comments())
6994            this = self.expression(exp.Alias, comments=comments, this=this, alias=alias)
6995            column = this.this
6996
6997            # Moves the comment next to the alias in `expr /* comment */ AS alias`
6998            if not this.comments and column and column.comments:
6999                this.comments = column.pop_comments()
7000
7001        return this
7002
7003    def _parse_id_var(
7004        self,
7005        any_token: bool = True,
7006        tokens: t.Optional[t.Collection[TokenType]] = None,
7007    ) -> t.Optional[exp.Expression]:
7008        expression = self._parse_identifier()
7009        if not expression and (
7010            (any_token and self._advance_any()) or self._match_set(tokens or self.ID_VAR_TOKENS)
7011        ):
7012            quoted = self._prev.token_type == TokenType.STRING
7013            expression = self._identifier_expression(quoted=quoted)
7014
7015        return expression
7016
7017    def _parse_string(self) -> t.Optional[exp.Expression]:
7018        if self._match_set(self.STRING_PARSERS):
7019            return self.STRING_PARSERS[self._prev.token_type](self, self._prev)
7020        return self._parse_placeholder()
7021
7022    def _parse_string_as_identifier(self) -> t.Optional[exp.Identifier]:
7023        output = exp.to_identifier(self._match(TokenType.STRING) and self._prev.text, quoted=True)
7024        if output:
7025            output.update_positions(self._prev)
7026        return output
7027
7028    def _parse_number(self) -> t.Optional[exp.Expression]:
7029        if self._match_set(self.NUMERIC_PARSERS):
7030            return self.NUMERIC_PARSERS[self._prev.token_type](self, self._prev)
7031        return self._parse_placeholder()
7032
7033    def _parse_identifier(self) -> t.Optional[exp.Expression]:
7034        if self._match(TokenType.IDENTIFIER):
7035            return self._identifier_expression(quoted=True)
7036        return self._parse_placeholder()
7037
7038    def _parse_var(
7039        self,
7040        any_token: bool = False,
7041        tokens: t.Optional[t.Collection[TokenType]] = None,
7042        upper: bool = False,
7043    ) -> t.Optional[exp.Expression]:
7044        if (
7045            (any_token and self._advance_any())
7046            or self._match(TokenType.VAR)
7047            or (self._match_set(tokens) if tokens else False)
7048        ):
7049            return self.expression(
7050                exp.Var, this=self._prev.text.upper() if upper else self._prev.text
7051            )
7052        return self._parse_placeholder()
7053
7054    def _advance_any(self, ignore_reserved: bool = False) -> t.Optional[Token]:
7055        if self._curr and (ignore_reserved or self._curr.token_type not in self.RESERVED_TOKENS):
7056            self._advance()
7057            return self._prev
7058        return None
7059
7060    def _parse_var_or_string(self, upper: bool = False) -> t.Optional[exp.Expression]:
7061        return self._parse_string() or self._parse_var(any_token=True, upper=upper)
7062
7063    def _parse_primary_or_var(self) -> t.Optional[exp.Expression]:
7064        return self._parse_primary() or self._parse_var(any_token=True)
7065
7066    def _parse_null(self) -> t.Optional[exp.Expression]:
7067        if self._match_set(self.NULL_TOKENS):
7068            return self.PRIMARY_PARSERS[TokenType.NULL](self, self._prev)
7069        return self._parse_placeholder()
7070
7071    def _parse_boolean(self) -> t.Optional[exp.Expression]:
7072        if self._match(TokenType.TRUE):
7073            return self.PRIMARY_PARSERS[TokenType.TRUE](self, self._prev)
7074        if self._match(TokenType.FALSE):
7075            return self.PRIMARY_PARSERS[TokenType.FALSE](self, self._prev)
7076        return self._parse_placeholder()
7077
7078    def _parse_star(self) -> t.Optional[exp.Expression]:
7079        if self._match(TokenType.STAR):
7080            return self.PRIMARY_PARSERS[TokenType.STAR](self, self._prev)
7081        return self._parse_placeholder()
7082
7083    def _parse_parameter(self) -> exp.Parameter:
7084        this = self._parse_identifier() or self._parse_primary_or_var()
7085        return self.expression(exp.Parameter, this=this)
7086
7087    def _parse_placeholder(self) -> t.Optional[exp.Expression]:
7088        if self._match_set(self.PLACEHOLDER_PARSERS):
7089            placeholder = self.PLACEHOLDER_PARSERS[self._prev.token_type](self)
7090            if placeholder:
7091                return placeholder
7092            self._advance(-1)
7093        return None
7094
7095    def _parse_star_op(self, *keywords: str) -> t.Optional[t.List[exp.Expression]]:
7096        if not self._match_texts(keywords):
7097            return None
7098        if self._match(TokenType.L_PAREN, advance=False):
7099            return self._parse_wrapped_csv(self._parse_expression)
7100
7101        expression = self._parse_expression()
7102        return [expression] if expression else None
7103
7104    def _parse_csv(
7105        self, parse_method: t.Callable, sep: TokenType = TokenType.COMMA
7106    ) -> t.List[exp.Expression]:
7107        parse_result = parse_method()
7108        items = [parse_result] if parse_result is not None else []
7109
7110        while self._match(sep):
7111            self._add_comments(parse_result)
7112            parse_result = parse_method()
7113            if parse_result is not None:
7114                items.append(parse_result)
7115
7116        return items
7117
7118    def _parse_tokens(
7119        self, parse_method: t.Callable, expressions: t.Dict
7120    ) -> t.Optional[exp.Expression]:
7121        this = parse_method()
7122
7123        while self._match_set(expressions):
7124            this = self.expression(
7125                expressions[self._prev.token_type],
7126                this=this,
7127                comments=self._prev_comments,
7128                expression=parse_method(),
7129            )
7130
7131        return this
7132
7133    def _parse_wrapped_id_vars(self, optional: bool = False) -> t.List[exp.Expression]:
7134        return self._parse_wrapped_csv(self._parse_id_var, optional=optional)
7135
7136    def _parse_wrapped_csv(
7137        self, parse_method: t.Callable, sep: TokenType = TokenType.COMMA, optional: bool = False
7138    ) -> t.List[exp.Expression]:
7139        return self._parse_wrapped(
7140            lambda: self._parse_csv(parse_method, sep=sep), optional=optional
7141        )
7142
7143    def _parse_wrapped(self, parse_method: t.Callable, optional: bool = False) -> t.Any:
7144        wrapped = self._match(TokenType.L_PAREN)
7145        if not wrapped and not optional:
7146            self.raise_error("Expecting (")
7147        parse_result = parse_method()
7148        if wrapped:
7149            self._match_r_paren()
7150        return parse_result
7151
7152    def _parse_expressions(self) -> t.List[exp.Expression]:
7153        return self._parse_csv(self._parse_expression)
7154
7155    def _parse_select_or_expression(self, alias: bool = False) -> t.Optional[exp.Expression]:
7156        return self._parse_select() or self._parse_set_operations(
7157            self._parse_alias(self._parse_assignment(), explicit=True)
7158            if alias
7159            else self._parse_assignment()
7160        )
7161
7162    def _parse_ddl_select(self) -> t.Optional[exp.Expression]:
7163        return self._parse_query_modifiers(
7164            self._parse_set_operations(self._parse_select(nested=True, parse_subquery_alias=False))
7165        )
7166
7167    def _parse_transaction(self) -> exp.Transaction | exp.Command:
7168        this = None
7169        if self._match_texts(self.TRANSACTION_KIND):
7170            this = self._prev.text
7171
7172        self._match_texts(("TRANSACTION", "WORK"))
7173
7174        modes = []
7175        while True:
7176            mode = []
7177            while self._match(TokenType.VAR):
7178                mode.append(self._prev.text)
7179
7180            if mode:
7181                modes.append(" ".join(mode))
7182            if not self._match(TokenType.COMMA):
7183                break
7184
7185        return self.expression(exp.Transaction, this=this, modes=modes)
7186
7187    def _parse_commit_or_rollback(self) -> exp.Commit | exp.Rollback:
7188        chain = None
7189        savepoint = None
7190        is_rollback = self._prev.token_type == TokenType.ROLLBACK
7191
7192        self._match_texts(("TRANSACTION", "WORK"))
7193
7194        if self._match_text_seq("TO"):
7195            self._match_text_seq("SAVEPOINT")
7196            savepoint = self._parse_id_var()
7197
7198        if self._match(TokenType.AND):
7199            chain = not self._match_text_seq("NO")
7200            self._match_text_seq("CHAIN")
7201
7202        if is_rollback:
7203            return self.expression(exp.Rollback, savepoint=savepoint)
7204
7205        return self.expression(exp.Commit, chain=chain)
7206
7207    def _parse_refresh(self) -> exp.Refresh:
7208        self._match(TokenType.TABLE)
7209        return self.expression(exp.Refresh, this=self._parse_string() or self._parse_table())
7210
7211    def _parse_add_column(self) -> t.Optional[exp.Expression]:
7212        if not self._match_text_seq("ADD"):
7213            return None
7214
7215        self._match(TokenType.COLUMN)
7216        exists_column = self._parse_exists(not_=True)
7217        expression = self._parse_field_def()
7218
7219        if expression:
7220            expression.set("exists", exists_column)
7221
7222            # https://docs.databricks.com/delta/update-schema.html#explicitly-update-schema-to-add-columns
7223            if self._match_texts(("FIRST", "AFTER")):
7224                position = self._prev.text
7225                column_position = self.expression(
7226                    exp.ColumnPosition, this=self._parse_column(), position=position
7227                )
7228                expression.set("position", column_position)
7229
7230        return expression
7231
7232    def _parse_drop_column(self) -> t.Optional[exp.Drop | exp.Command]:
7233        drop = self._match(TokenType.DROP) and self._parse_drop()
7234        if drop and not isinstance(drop, exp.Command):
7235            drop.set("kind", drop.args.get("kind", "COLUMN"))
7236        return drop
7237
7238    # https://docs.aws.amazon.com/athena/latest/ug/alter-table-drop-partition.html
7239    def _parse_drop_partition(self, exists: t.Optional[bool] = None) -> exp.DropPartition:
7240        return self.expression(
7241            exp.DropPartition, expressions=self._parse_csv(self._parse_partition), exists=exists
7242        )
7243
7244    def _parse_alter_table_add(self) -> t.List[exp.Expression]:
7245        index = self._index - 1
7246
7247        if self._match_set(self.ADD_CONSTRAINT_TOKENS, advance=False):
7248            return self._parse_csv(
7249                lambda: self.expression(
7250                    exp.AddConstraint, expressions=self._parse_csv(self._parse_constraint)
7251                )
7252            )
7253
7254        self._retreat(index)
7255        if not self.ALTER_TABLE_ADD_REQUIRED_FOR_EACH_COLUMN and self._match_text_seq("ADD"):
7256            return self._parse_wrapped_csv(self._parse_field_def, optional=True)
7257
7258        if self._match_text_seq("ADD", "COLUMNS"):
7259            schema = self._parse_schema()
7260            if schema:
7261                return [schema]
7262            return []
7263
7264        return self._parse_wrapped_csv(self._parse_add_column, optional=True)
7265
7266    def _parse_alter_table_alter(self) -> t.Optional[exp.Expression]:
7267        if self._match_texts(self.ALTER_ALTER_PARSERS):
7268            return self.ALTER_ALTER_PARSERS[self._prev.text.upper()](self)
7269
7270        # Many dialects support the ALTER [COLUMN] syntax, so if there is no
7271        # keyword after ALTER we default to parsing this statement
7272        self._match(TokenType.COLUMN)
7273        column = self._parse_field(any_token=True)
7274
7275        if self._match_pair(TokenType.DROP, TokenType.DEFAULT):
7276            return self.expression(exp.AlterColumn, this=column, drop=True)
7277        if self._match_pair(TokenType.SET, TokenType.DEFAULT):
7278            return self.expression(exp.AlterColumn, this=column, default=self._parse_assignment())
7279        if self._match(TokenType.COMMENT):
7280            return self.expression(exp.AlterColumn, this=column, comment=self._parse_string())
7281        if self._match_text_seq("DROP", "NOT", "NULL"):
7282            return self.expression(
7283                exp.AlterColumn,
7284                this=column,
7285                drop=True,
7286                allow_null=True,
7287            )
7288        if self._match_text_seq("SET", "NOT", "NULL"):
7289            return self.expression(
7290                exp.AlterColumn,
7291                this=column,
7292                allow_null=False,
7293            )
7294
7295        if self._match_text_seq("SET", "VISIBLE"):
7296            return self.expression(exp.AlterColumn, this=column, visible="VISIBLE")
7297        if self._match_text_seq("SET", "INVISIBLE"):
7298            return self.expression(exp.AlterColumn, this=column, visible="INVISIBLE")
7299
7300        self._match_text_seq("SET", "DATA")
7301        self._match_text_seq("TYPE")
7302        return self.expression(
7303            exp.AlterColumn,
7304            this=column,
7305            dtype=self._parse_types(),
7306            collate=self._match(TokenType.COLLATE) and self._parse_term(),
7307            using=self._match(TokenType.USING) and self._parse_assignment(),
7308        )
7309
7310    def _parse_alter_diststyle(self) -> exp.AlterDistStyle:
7311        if self._match_texts(("ALL", "EVEN", "AUTO")):
7312            return self.expression(exp.AlterDistStyle, this=exp.var(self._prev.text.upper()))
7313
7314        self._match_text_seq("KEY", "DISTKEY")
7315        return self.expression(exp.AlterDistStyle, this=self._parse_column())
7316
7317    def _parse_alter_sortkey(self, compound: t.Optional[bool] = None) -> exp.AlterSortKey:
7318        if compound:
7319            self._match_text_seq("SORTKEY")
7320
7321        if self._match(TokenType.L_PAREN, advance=False):
7322            return self.expression(
7323                exp.AlterSortKey, expressions=self._parse_wrapped_id_vars(), compound=compound
7324            )
7325
7326        self._match_texts(("AUTO", "NONE"))
7327        return self.expression(
7328            exp.AlterSortKey, this=exp.var(self._prev.text.upper()), compound=compound
7329        )
7330
7331    def _parse_alter_table_drop(self) -> t.List[exp.Expression]:
7332        index = self._index - 1
7333
7334        partition_exists = self._parse_exists()
7335        if self._match(TokenType.PARTITION, advance=False):
7336            return self._parse_csv(lambda: self._parse_drop_partition(exists=partition_exists))
7337
7338        self._retreat(index)
7339        return self._parse_csv(self._parse_drop_column)
7340
7341    def _parse_alter_table_rename(self) -> t.Optional[exp.AlterRename | exp.RenameColumn]:
7342        if self._match(TokenType.COLUMN):
7343            exists = self._parse_exists()
7344            old_column = self._parse_column()
7345            to = self._match_text_seq("TO")
7346            new_column = self._parse_column()
7347
7348            if old_column is None or to is None or new_column is None:
7349                return None
7350
7351            return self.expression(exp.RenameColumn, this=old_column, to=new_column, exists=exists)
7352
7353        self._match_text_seq("TO")
7354        return self.expression(exp.AlterRename, this=self._parse_table(schema=True))
7355
7356    def _parse_alter_table_set(self) -> exp.AlterSet:
7357        alter_set = self.expression(exp.AlterSet)
7358
7359        if self._match(TokenType.L_PAREN, advance=False) or self._match_text_seq(
7360            "TABLE", "PROPERTIES"
7361        ):
7362            alter_set.set("expressions", self._parse_wrapped_csv(self._parse_assignment))
7363        elif self._match_text_seq("FILESTREAM_ON", advance=False):
7364            alter_set.set("expressions", [self._parse_assignment()])
7365        elif self._match_texts(("LOGGED", "UNLOGGED")):
7366            alter_set.set("option", exp.var(self._prev.text.upper()))
7367        elif self._match_text_seq("WITHOUT") and self._match_texts(("CLUSTER", "OIDS")):
7368            alter_set.set("option", exp.var(f"WITHOUT {self._prev.text.upper()}"))
7369        elif self._match_text_seq("LOCATION"):
7370            alter_set.set("location", self._parse_field())
7371        elif self._match_text_seq("ACCESS", "METHOD"):
7372            alter_set.set("access_method", self._parse_field())
7373        elif self._match_text_seq("TABLESPACE"):
7374            alter_set.set("tablespace", self._parse_field())
7375        elif self._match_text_seq("FILE", "FORMAT") or self._match_text_seq("FILEFORMAT"):
7376            alter_set.set("file_format", [self._parse_field()])
7377        elif self._match_text_seq("STAGE_FILE_FORMAT"):
7378            alter_set.set("file_format", self._parse_wrapped_options())
7379        elif self._match_text_seq("STAGE_COPY_OPTIONS"):
7380            alter_set.set("copy_options", self._parse_wrapped_options())
7381        elif self._match_text_seq("TAG") or self._match_text_seq("TAGS"):
7382            alter_set.set("tag", self._parse_csv(self._parse_assignment))
7383        else:
7384            if self._match_text_seq("SERDE"):
7385                alter_set.set("serde", self._parse_field())
7386
7387            alter_set.set("expressions", [self._parse_properties()])
7388
7389        return alter_set
7390
7391    def _parse_alter(self) -> exp.Alter | exp.Command:
7392        start = self._prev
7393
7394        alter_token = self._match_set(self.ALTERABLES) and self._prev
7395        if not alter_token:
7396            return self._parse_as_command(start)
7397
7398        exists = self._parse_exists()
7399        only = self._match_text_seq("ONLY")
7400        this = self._parse_table(schema=True)
7401        cluster = self._parse_on_property() if self._match(TokenType.ON) else None
7402
7403        if self._next:
7404            self._advance()
7405
7406        parser = self.ALTER_PARSERS.get(self._prev.text.upper()) if self._prev else None
7407        if parser:
7408            actions = ensure_list(parser(self))
7409            not_valid = self._match_text_seq("NOT", "VALID")
7410            options = self._parse_csv(self._parse_property)
7411
7412            if not self._curr and actions:
7413                return self.expression(
7414                    exp.Alter,
7415                    this=this,
7416                    kind=alter_token.text.upper(),
7417                    exists=exists,
7418                    actions=actions,
7419                    only=only,
7420                    options=options,
7421                    cluster=cluster,
7422                    not_valid=not_valid,
7423                )
7424
7425        return self._parse_as_command(start)
7426
7427    def _parse_analyze(self) -> exp.Analyze | exp.Command:
7428        start = self._prev
7429        # https://duckdb.org/docs/sql/statements/analyze
7430        if not self._curr:
7431            return self.expression(exp.Analyze)
7432
7433        options = []
7434        while self._match_texts(self.ANALYZE_STYLES):
7435            if self._prev.text.upper() == "BUFFER_USAGE_LIMIT":
7436                options.append(f"BUFFER_USAGE_LIMIT {self._parse_number()}")
7437            else:
7438                options.append(self._prev.text.upper())
7439
7440        this: t.Optional[exp.Expression] = None
7441        inner_expression: t.Optional[exp.Expression] = None
7442
7443        kind = self._curr and self._curr.text.upper()
7444
7445        if self._match(TokenType.TABLE) or self._match(TokenType.INDEX):
7446            this = self._parse_table_parts()
7447        elif self._match_text_seq("TABLES"):
7448            if self._match_set((TokenType.FROM, TokenType.IN)):
7449                kind = f"{kind} {self._prev.text.upper()}"
7450                this = self._parse_table(schema=True, is_db_reference=True)
7451        elif self._match_text_seq("DATABASE"):
7452            this = self._parse_table(schema=True, is_db_reference=True)
7453        elif self._match_text_seq("CLUSTER"):
7454            this = self._parse_table()
7455        # Try matching inner expr keywords before fallback to parse table.
7456        elif self._match_texts(self.ANALYZE_EXPRESSION_PARSERS):
7457            kind = None
7458            inner_expression = self.ANALYZE_EXPRESSION_PARSERS[self._prev.text.upper()](self)
7459        else:
7460            # Empty kind  https://prestodb.io/docs/current/sql/analyze.html
7461            kind = None
7462            this = self._parse_table_parts()
7463
7464        partition = self._try_parse(self._parse_partition)
7465        if not partition and self._match_texts(self.PARTITION_KEYWORDS):
7466            return self._parse_as_command(start)
7467
7468        # https://docs.starrocks.io/docs/sql-reference/sql-statements/cbo_stats/ANALYZE_TABLE/
7469        if self._match_text_seq("WITH", "SYNC", "MODE") or self._match_text_seq(
7470            "WITH", "ASYNC", "MODE"
7471        ):
7472            mode = f"WITH {self._tokens[self._index - 2].text.upper()} MODE"
7473        else:
7474            mode = None
7475
7476        if self._match_texts(self.ANALYZE_EXPRESSION_PARSERS):
7477            inner_expression = self.ANALYZE_EXPRESSION_PARSERS[self._prev.text.upper()](self)
7478
7479        properties = self._parse_properties()
7480        return self.expression(
7481            exp.Analyze,
7482            kind=kind,
7483            this=this,
7484            mode=mode,
7485            partition=partition,
7486            properties=properties,
7487            expression=inner_expression,
7488            options=options,
7489        )
7490
7491    # https://spark.apache.org/docs/3.5.1/sql-ref-syntax-aux-analyze-table.html
7492    def _parse_analyze_statistics(self) -> exp.AnalyzeStatistics:
7493        this = None
7494        kind = self._prev.text.upper()
7495        option = self._prev.text.upper() if self._match_text_seq("DELTA") else None
7496        expressions = []
7497
7498        if not self._match_text_seq("STATISTICS"):
7499            self.raise_error("Expecting token STATISTICS")
7500
7501        if self._match_text_seq("NOSCAN"):
7502            this = "NOSCAN"
7503        elif self._match(TokenType.FOR):
7504            if self._match_text_seq("ALL", "COLUMNS"):
7505                this = "FOR ALL COLUMNS"
7506            if self._match_texts("COLUMNS"):
7507                this = "FOR COLUMNS"
7508                expressions = self._parse_csv(self._parse_column_reference)
7509        elif self._match_text_seq("SAMPLE"):
7510            sample = self._parse_number()
7511            expressions = [
7512                self.expression(
7513                    exp.AnalyzeSample,
7514                    sample=sample,
7515                    kind=self._prev.text.upper() if self._match(TokenType.PERCENT) else None,
7516                )
7517            ]
7518
7519        return self.expression(
7520            exp.AnalyzeStatistics, kind=kind, option=option, this=this, expressions=expressions
7521        )
7522
7523    # https://docs.oracle.com/en/database/oracle/oracle-database/21/sqlrf/ANALYZE.html
7524    def _parse_analyze_validate(self) -> exp.AnalyzeValidate:
7525        kind = None
7526        this = None
7527        expression: t.Optional[exp.Expression] = None
7528        if self._match_text_seq("REF", "UPDATE"):
7529            kind = "REF"
7530            this = "UPDATE"
7531            if self._match_text_seq("SET", "DANGLING", "TO", "NULL"):
7532                this = "UPDATE SET DANGLING TO NULL"
7533        elif self._match_text_seq("STRUCTURE"):
7534            kind = "STRUCTURE"
7535            if self._match_text_seq("CASCADE", "FAST"):
7536                this = "CASCADE FAST"
7537            elif self._match_text_seq("CASCADE", "COMPLETE") and self._match_texts(
7538                ("ONLINE", "OFFLINE")
7539            ):
7540                this = f"CASCADE COMPLETE {self._prev.text.upper()}"
7541                expression = self._parse_into()
7542
7543        return self.expression(exp.AnalyzeValidate, kind=kind, this=this, expression=expression)
7544
7545    def _parse_analyze_columns(self) -> t.Optional[exp.AnalyzeColumns]:
7546        this = self._prev.text.upper()
7547        if self._match_text_seq("COLUMNS"):
7548            return self.expression(exp.AnalyzeColumns, this=f"{this} {self._prev.text.upper()}")
7549        return None
7550
7551    def _parse_analyze_delete(self) -> t.Optional[exp.AnalyzeDelete]:
7552        kind = self._prev.text.upper() if self._match_text_seq("SYSTEM") else None
7553        if self._match_text_seq("STATISTICS"):
7554            return self.expression(exp.AnalyzeDelete, kind=kind)
7555        return None
7556
7557    def _parse_analyze_list(self) -> t.Optional[exp.AnalyzeListChainedRows]:
7558        if self._match_text_seq("CHAINED", "ROWS"):
7559            return self.expression(exp.AnalyzeListChainedRows, expression=self._parse_into())
7560        return None
7561
7562    # https://dev.mysql.com/doc/refman/8.4/en/analyze-table.html
7563    def _parse_analyze_histogram(self) -> exp.AnalyzeHistogram:
7564        this = self._prev.text.upper()
7565        expression: t.Optional[exp.Expression] = None
7566        expressions = []
7567        update_options = None
7568
7569        if self._match_text_seq("HISTOGRAM", "ON"):
7570            expressions = self._parse_csv(self._parse_column_reference)
7571            with_expressions = []
7572            while self._match(TokenType.WITH):
7573                # https://docs.starrocks.io/docs/sql-reference/sql-statements/cbo_stats/ANALYZE_TABLE/
7574                if self._match_texts(("SYNC", "ASYNC")):
7575                    if self._match_text_seq("MODE", advance=False):
7576                        with_expressions.append(f"{self._prev.text.upper()} MODE")
7577                        self._advance()
7578                else:
7579                    buckets = self._parse_number()
7580                    if self._match_text_seq("BUCKETS"):
7581                        with_expressions.append(f"{buckets} BUCKETS")
7582            if with_expressions:
7583                expression = self.expression(exp.AnalyzeWith, expressions=with_expressions)
7584
7585            if self._match_texts(("MANUAL", "AUTO")) and self._match(
7586                TokenType.UPDATE, advance=False
7587            ):
7588                update_options = self._prev.text.upper()
7589                self._advance()
7590            elif self._match_text_seq("USING", "DATA"):
7591                expression = self.expression(exp.UsingData, this=self._parse_string())
7592
7593        return self.expression(
7594            exp.AnalyzeHistogram,
7595            this=this,
7596            expressions=expressions,
7597            expression=expression,
7598            update_options=update_options,
7599        )
7600
7601    def _parse_merge(self) -> exp.Merge:
7602        self._match(TokenType.INTO)
7603        target = self._parse_table()
7604
7605        if target and self._match(TokenType.ALIAS, advance=False):
7606            target.set("alias", self._parse_table_alias())
7607
7608        self._match(TokenType.USING)
7609        using = self._parse_table()
7610
7611        self._match(TokenType.ON)
7612        on = self._parse_assignment()
7613
7614        return self.expression(
7615            exp.Merge,
7616            this=target,
7617            using=using,
7618            on=on,
7619            whens=self._parse_when_matched(),
7620            returning=self._parse_returning(),
7621        )
7622
7623    def _parse_when_matched(self) -> exp.Whens:
7624        whens = []
7625
7626        while self._match(TokenType.WHEN):
7627            matched = not self._match(TokenType.NOT)
7628            self._match_text_seq("MATCHED")
7629            source = (
7630                False
7631                if self._match_text_seq("BY", "TARGET")
7632                else self._match_text_seq("BY", "SOURCE")
7633            )
7634            condition = self._parse_assignment() if self._match(TokenType.AND) else None
7635
7636            self._match(TokenType.THEN)
7637
7638            if self._match(TokenType.INSERT):
7639                this = self._parse_star()
7640                if this:
7641                    then: t.Optional[exp.Expression] = self.expression(exp.Insert, this=this)
7642                else:
7643                    then = self.expression(
7644                        exp.Insert,
7645                        this=exp.var("ROW")
7646                        if self._match_text_seq("ROW")
7647                        else self._parse_value(values=False),
7648                        expression=self._match_text_seq("VALUES") and self._parse_value(),
7649                    )
7650            elif self._match(TokenType.UPDATE):
7651                expressions = self._parse_star()
7652                if expressions:
7653                    then = self.expression(exp.Update, expressions=expressions)
7654                else:
7655                    then = self.expression(
7656                        exp.Update,
7657                        expressions=self._match(TokenType.SET)
7658                        and self._parse_csv(self._parse_equality),
7659                    )
7660            elif self._match(TokenType.DELETE):
7661                then = self.expression(exp.Var, this=self._prev.text)
7662            else:
7663                then = self._parse_var_from_options(self.CONFLICT_ACTIONS)
7664
7665            whens.append(
7666                self.expression(
7667                    exp.When,
7668                    matched=matched,
7669                    source=source,
7670                    condition=condition,
7671                    then=then,
7672                )
7673            )
7674        return self.expression(exp.Whens, expressions=whens)
7675
7676    def _parse_show(self) -> t.Optional[exp.Expression]:
7677        parser = self._find_parser(self.SHOW_PARSERS, self.SHOW_TRIE)
7678        if parser:
7679            return parser(self)
7680        return self._parse_as_command(self._prev)
7681
7682    def _parse_set_item_assignment(
7683        self, kind: t.Optional[str] = None
7684    ) -> t.Optional[exp.Expression]:
7685        index = self._index
7686
7687        if kind in ("GLOBAL", "SESSION") and self._match_text_seq("TRANSACTION"):
7688            return self._parse_set_transaction(global_=kind == "GLOBAL")
7689
7690        left = self._parse_primary() or self._parse_column()
7691        assignment_delimiter = self._match_texts(("=", "TO"))
7692
7693        if not left or (self.SET_REQUIRES_ASSIGNMENT_DELIMITER and not assignment_delimiter):
7694            self._retreat(index)
7695            return None
7696
7697        right = self._parse_statement() or self._parse_id_var()
7698        if isinstance(right, (exp.Column, exp.Identifier)):
7699            right = exp.var(right.name)
7700
7701        this = self.expression(exp.EQ, this=left, expression=right)
7702        return self.expression(exp.SetItem, this=this, kind=kind)
7703
7704    def _parse_set_transaction(self, global_: bool = False) -> exp.Expression:
7705        self._match_text_seq("TRANSACTION")
7706        characteristics = self._parse_csv(
7707            lambda: self._parse_var_from_options(self.TRANSACTION_CHARACTERISTICS)
7708        )
7709        return self.expression(
7710            exp.SetItem,
7711            expressions=characteristics,
7712            kind="TRANSACTION",
7713            **{"global": global_},  # type: ignore
7714        )
7715
7716    def _parse_set_item(self) -> t.Optional[exp.Expression]:
7717        parser = self._find_parser(self.SET_PARSERS, self.SET_TRIE)
7718        return parser(self) if parser else self._parse_set_item_assignment(kind=None)
7719
7720    def _parse_set(self, unset: bool = False, tag: bool = False) -> exp.Set | exp.Command:
7721        index = self._index
7722        set_ = self.expression(
7723            exp.Set, expressions=self._parse_csv(self._parse_set_item), unset=unset, tag=tag
7724        )
7725
7726        if self._curr:
7727            self._retreat(index)
7728            return self._parse_as_command(self._prev)
7729
7730        return set_
7731
7732    def _parse_var_from_options(
7733        self, options: OPTIONS_TYPE, raise_unmatched: bool = True
7734    ) -> t.Optional[exp.Var]:
7735        start = self._curr
7736        if not start:
7737            return None
7738
7739        option = start.text.upper()
7740        continuations = options.get(option)
7741
7742        index = self._index
7743        self._advance()
7744        for keywords in continuations or []:
7745            if isinstance(keywords, str):
7746                keywords = (keywords,)
7747
7748            if self._match_text_seq(*keywords):
7749                option = f"{option} {' '.join(keywords)}"
7750                break
7751        else:
7752            if continuations or continuations is None:
7753                if raise_unmatched:
7754                    self.raise_error(f"Unknown option {option}")
7755
7756                self._retreat(index)
7757                return None
7758
7759        return exp.var(option)
7760
7761    def _parse_as_command(self, start: Token) -> exp.Command:
7762        while self._curr:
7763            self._advance()
7764        text = self._find_sql(start, self._prev)
7765        size = len(start.text)
7766        self._warn_unsupported()
7767        return exp.Command(this=text[:size], expression=text[size:])
7768
7769    def _parse_dict_property(self, this: str) -> exp.DictProperty:
7770        settings = []
7771
7772        self._match_l_paren()
7773        kind = self._parse_id_var()
7774
7775        if self._match(TokenType.L_PAREN):
7776            while True:
7777                key = self._parse_id_var()
7778                value = self._parse_primary()
7779                if not key and value is None:
7780                    break
7781                settings.append(self.expression(exp.DictSubProperty, this=key, value=value))
7782            self._match(TokenType.R_PAREN)
7783
7784        self._match_r_paren()
7785
7786        return self.expression(
7787            exp.DictProperty,
7788            this=this,
7789            kind=kind.this if kind else None,
7790            settings=settings,
7791        )
7792
7793    def _parse_dict_range(self, this: str) -> exp.DictRange:
7794        self._match_l_paren()
7795        has_min = self._match_text_seq("MIN")
7796        if has_min:
7797            min = self._parse_var() or self._parse_primary()
7798            self._match_text_seq("MAX")
7799            max = self._parse_var() or self._parse_primary()
7800        else:
7801            max = self._parse_var() or self._parse_primary()
7802            min = exp.Literal.number(0)
7803        self._match_r_paren()
7804        return self.expression(exp.DictRange, this=this, min=min, max=max)
7805
7806    def _parse_comprehension(
7807        self, this: t.Optional[exp.Expression]
7808    ) -> t.Optional[exp.Comprehension]:
7809        index = self._index
7810        expression = self._parse_column()
7811        if not self._match(TokenType.IN):
7812            self._retreat(index - 1)
7813            return None
7814        iterator = self._parse_column()
7815        condition = self._parse_assignment() if self._match_text_seq("IF") else None
7816        return self.expression(
7817            exp.Comprehension,
7818            this=this,
7819            expression=expression,
7820            iterator=iterator,
7821            condition=condition,
7822        )
7823
7824    def _parse_heredoc(self) -> t.Optional[exp.Heredoc]:
7825        if self._match(TokenType.HEREDOC_STRING):
7826            return self.expression(exp.Heredoc, this=self._prev.text)
7827
7828        if not self._match_text_seq("$"):
7829            return None
7830
7831        tags = ["$"]
7832        tag_text = None
7833
7834        if self._is_connected():
7835            self._advance()
7836            tags.append(self._prev.text.upper())
7837        else:
7838            self.raise_error("No closing $ found")
7839
7840        if tags[-1] != "$":
7841            if self._is_connected() and self._match_text_seq("$"):
7842                tag_text = tags[-1]
7843                tags.append("$")
7844            else:
7845                self.raise_error("No closing $ found")
7846
7847        heredoc_start = self._curr
7848
7849        while self._curr:
7850            if self._match_text_seq(*tags, advance=False):
7851                this = self._find_sql(heredoc_start, self._prev)
7852                self._advance(len(tags))
7853                return self.expression(exp.Heredoc, this=this, tag=tag_text)
7854
7855            self._advance()
7856
7857        self.raise_error(f"No closing {''.join(tags)} found")
7858        return None
7859
7860    def _find_parser(
7861        self, parsers: t.Dict[str, t.Callable], trie: t.Dict
7862    ) -> t.Optional[t.Callable]:
7863        if not self._curr:
7864            return None
7865
7866        index = self._index
7867        this = []
7868        while True:
7869            # The current token might be multiple words
7870            curr = self._curr.text.upper()
7871            key = curr.split(" ")
7872            this.append(curr)
7873
7874            self._advance()
7875            result, trie = in_trie(trie, key)
7876            if result == TrieResult.FAILED:
7877                break
7878
7879            if result == TrieResult.EXISTS:
7880                subparser = parsers[" ".join(this)]
7881                return subparser
7882
7883        self._retreat(index)
7884        return None
7885
7886    def _match(self, token_type, advance=True, expression=None):
7887        if not self._curr:
7888            return None
7889
7890        if self._curr.token_type == token_type:
7891            if advance:
7892                self._advance()
7893            self._add_comments(expression)
7894            return True
7895
7896        return None
7897
7898    def _match_set(self, types, advance=True):
7899        if not self._curr:
7900            return None
7901
7902        if self._curr.token_type in types:
7903            if advance:
7904                self._advance()
7905            return True
7906
7907        return None
7908
7909    def _match_pair(self, token_type_a, token_type_b, advance=True):
7910        if not self._curr or not self._next:
7911            return None
7912
7913        if self._curr.token_type == token_type_a and self._next.token_type == token_type_b:
7914            if advance:
7915                self._advance(2)
7916            return True
7917
7918        return None
7919
7920    def _match_l_paren(self, expression: t.Optional[exp.Expression] = None) -> None:
7921        if not self._match(TokenType.L_PAREN, expression=expression):
7922            self.raise_error("Expecting (")
7923
7924    def _match_r_paren(self, expression: t.Optional[exp.Expression] = None) -> None:
7925        if not self._match(TokenType.R_PAREN, expression=expression):
7926            self.raise_error("Expecting )")
7927
7928    def _match_texts(self, texts, advance=True):
7929        if (
7930            self._curr
7931            and self._curr.token_type != TokenType.STRING
7932            and self._curr.text.upper() in texts
7933        ):
7934            if advance:
7935                self._advance()
7936            return True
7937        return None
7938
7939    def _match_text_seq(self, *texts, advance=True):
7940        index = self._index
7941        for text in texts:
7942            if (
7943                self._curr
7944                and self._curr.token_type != TokenType.STRING
7945                and self._curr.text.upper() == text
7946            ):
7947                self._advance()
7948            else:
7949                self._retreat(index)
7950                return None
7951
7952        if not advance:
7953            self._retreat(index)
7954
7955        return True
7956
7957    def _replace_lambda(
7958        self, node: t.Optional[exp.Expression], expressions: t.List[exp.Expression]
7959    ) -> t.Optional[exp.Expression]:
7960        if not node:
7961            return node
7962
7963        lambda_types = {e.name: e.args.get("to") or False for e in expressions}
7964
7965        for column in node.find_all(exp.Column):
7966            typ = lambda_types.get(column.parts[0].name)
7967            if typ is not None:
7968                dot_or_id = column.to_dot() if column.table else column.this
7969
7970                if typ:
7971                    dot_or_id = self.expression(
7972                        exp.Cast,
7973                        this=dot_or_id,
7974                        to=typ,
7975                    )
7976
7977                parent = column.parent
7978
7979                while isinstance(parent, exp.Dot):
7980                    if not isinstance(parent.parent, exp.Dot):
7981                        parent.replace(dot_or_id)
7982                        break
7983                    parent = parent.parent
7984                else:
7985                    if column is node:
7986                        node = dot_or_id
7987                    else:
7988                        column.replace(dot_or_id)
7989        return node
7990
7991    def _parse_truncate_table(self) -> t.Optional[exp.TruncateTable] | exp.Expression:
7992        start = self._prev
7993
7994        # Not to be confused with TRUNCATE(number, decimals) function call
7995        if self._match(TokenType.L_PAREN):
7996            self._retreat(self._index - 2)
7997            return self._parse_function()
7998
7999        # Clickhouse supports TRUNCATE DATABASE as well
8000        is_database = self._match(TokenType.DATABASE)
8001
8002        self._match(TokenType.TABLE)
8003
8004        exists = self._parse_exists(not_=False)
8005
8006        expressions = self._parse_csv(
8007            lambda: self._parse_table(schema=True, is_db_reference=is_database)
8008        )
8009
8010        cluster = self._parse_on_property() if self._match(TokenType.ON) else None
8011
8012        if self._match_text_seq("RESTART", "IDENTITY"):
8013            identity = "RESTART"
8014        elif self._match_text_seq("CONTINUE", "IDENTITY"):
8015            identity = "CONTINUE"
8016        else:
8017            identity = None
8018
8019        if self._match_text_seq("CASCADE") or self._match_text_seq("RESTRICT"):
8020            option = self._prev.text
8021        else:
8022            option = None
8023
8024        partition = self._parse_partition()
8025
8026        # Fallback case
8027        if self._curr:
8028            return self._parse_as_command(start)
8029
8030        return self.expression(
8031            exp.TruncateTable,
8032            expressions=expressions,
8033            is_database=is_database,
8034            exists=exists,
8035            cluster=cluster,
8036            identity=identity,
8037            option=option,
8038            partition=partition,
8039        )
8040
8041    def _parse_with_operator(self) -> t.Optional[exp.Expression]:
8042        this = self._parse_ordered(self._parse_opclass)
8043
8044        if not self._match(TokenType.WITH):
8045            return this
8046
8047        op = self._parse_var(any_token=True)
8048
8049        return self.expression(exp.WithOperator, this=this, op=op)
8050
8051    def _parse_wrapped_options(self) -> t.List[t.Optional[exp.Expression]]:
8052        self._match(TokenType.EQ)
8053        self._match(TokenType.L_PAREN)
8054
8055        opts: t.List[t.Optional[exp.Expression]] = []
8056        option: exp.Expression | None
8057        while self._curr and not self._match(TokenType.R_PAREN):
8058            if self._match_text_seq("FORMAT_NAME", "="):
8059                # The FORMAT_NAME can be set to an identifier for Snowflake and T-SQL
8060                option = self._parse_format_name()
8061            else:
8062                option = self._parse_property()
8063
8064            if option is None:
8065                self.raise_error("Unable to parse option")
8066                break
8067
8068            opts.append(option)
8069
8070        return opts
8071
8072    def _parse_copy_parameters(self) -> t.List[exp.CopyParameter]:
8073        sep = TokenType.COMMA if self.dialect.COPY_PARAMS_ARE_CSV else None
8074
8075        options = []
8076        while self._curr and not self._match(TokenType.R_PAREN, advance=False):
8077            option = self._parse_var(any_token=True)
8078            prev = self._prev.text.upper()
8079
8080            # Different dialects might separate options and values by white space, "=" and "AS"
8081            self._match(TokenType.EQ)
8082            self._match(TokenType.ALIAS)
8083
8084            param = self.expression(exp.CopyParameter, this=option)
8085
8086            if prev in self.COPY_INTO_VARLEN_OPTIONS and self._match(
8087                TokenType.L_PAREN, advance=False
8088            ):
8089                # Snowflake FILE_FORMAT case, Databricks COPY & FORMAT options
8090                param.set("expressions", self._parse_wrapped_options())
8091            elif prev == "FILE_FORMAT":
8092                # T-SQL's external file format case
8093                param.set("expression", self._parse_field())
8094            else:
8095                param.set("expression", self._parse_unquoted_field())
8096
8097            options.append(param)
8098            self._match(sep)
8099
8100        return options
8101
8102    def _parse_credentials(self) -> t.Optional[exp.Credentials]:
8103        expr = self.expression(exp.Credentials)
8104
8105        if self._match_text_seq("STORAGE_INTEGRATION", "="):
8106            expr.set("storage", self._parse_field())
8107        if self._match_text_seq("CREDENTIALS"):
8108            # Snowflake case: CREDENTIALS = (...), Redshift case: CREDENTIALS <string>
8109            creds = (
8110                self._parse_wrapped_options() if self._match(TokenType.EQ) else self._parse_field()
8111            )
8112            expr.set("credentials", creds)
8113        if self._match_text_seq("ENCRYPTION"):
8114            expr.set("encryption", self._parse_wrapped_options())
8115        if self._match_text_seq("IAM_ROLE"):
8116            expr.set("iam_role", self._parse_field())
8117        if self._match_text_seq("REGION"):
8118            expr.set("region", self._parse_field())
8119
8120        return expr
8121
8122    def _parse_file_location(self) -> t.Optional[exp.Expression]:
8123        return self._parse_field()
8124
8125    def _parse_copy(self) -> exp.Copy | exp.Command:
8126        start = self._prev
8127
8128        self._match(TokenType.INTO)
8129
8130        this = (
8131            self._parse_select(nested=True, parse_subquery_alias=False)
8132            if self._match(TokenType.L_PAREN, advance=False)
8133            else self._parse_table(schema=True)
8134        )
8135
8136        kind = self._match(TokenType.FROM) or not self._match_text_seq("TO")
8137
8138        files = self._parse_csv(self._parse_file_location)
8139        credentials = self._parse_credentials()
8140
8141        self._match_text_seq("WITH")
8142
8143        params = self._parse_wrapped(self._parse_copy_parameters, optional=True)
8144
8145        # Fallback case
8146        if self._curr:
8147            return self._parse_as_command(start)
8148
8149        return self.expression(
8150            exp.Copy,
8151            this=this,
8152            kind=kind,
8153            credentials=credentials,
8154            files=files,
8155            params=params,
8156        )
8157
8158    def _parse_normalize(self) -> exp.Normalize:
8159        return self.expression(
8160            exp.Normalize,
8161            this=self._parse_bitwise(),
8162            form=self._match(TokenType.COMMA) and self._parse_var(),
8163        )
8164
8165    def _parse_ceil_floor(self, expr_type: t.Type[TCeilFloor]) -> TCeilFloor:
8166        args = self._parse_csv(lambda: self._parse_lambda())
8167
8168        this = seq_get(args, 0)
8169        decimals = seq_get(args, 1)
8170
8171        return expr_type(
8172            this=this, decimals=decimals, to=self._match_text_seq("TO") and self._parse_var()
8173        )
8174
8175    def _parse_star_ops(self) -> t.Optional[exp.Expression]:
8176        if self._match_text_seq("COLUMNS", "(", advance=False):
8177            this = self._parse_function()
8178            if isinstance(this, exp.Columns):
8179                this.set("unpack", True)
8180            return this
8181
8182        return self.expression(
8183            exp.Star,
8184            **{  # type: ignore
8185                "except": self._parse_star_op("EXCEPT", "EXCLUDE"),
8186                "replace": self._parse_star_op("REPLACE"),
8187                "rename": self._parse_star_op("RENAME"),
8188            },
8189        )
8190
8191    def _parse_grant_privilege(self) -> t.Optional[exp.GrantPrivilege]:
8192        privilege_parts = []
8193
8194        # Keep consuming consecutive keywords until comma (end of this privilege) or ON
8195        # (end of privilege list) or L_PAREN (start of column list) are met
8196        while self._curr and not self._match_set(self.PRIVILEGE_FOLLOW_TOKENS, advance=False):
8197            privilege_parts.append(self._curr.text.upper())
8198            self._advance()
8199
8200        this = exp.var(" ".join(privilege_parts))
8201        expressions = (
8202            self._parse_wrapped_csv(self._parse_column)
8203            if self._match(TokenType.L_PAREN, advance=False)
8204            else None
8205        )
8206
8207        return self.expression(exp.GrantPrivilege, this=this, expressions=expressions)
8208
8209    def _parse_grant_principal(self) -> t.Optional[exp.GrantPrincipal]:
8210        kind = self._match_texts(("ROLE", "GROUP")) and self._prev.text.upper()
8211        principal = self._parse_id_var()
8212
8213        if not principal:
8214            return None
8215
8216        return self.expression(exp.GrantPrincipal, this=principal, kind=kind)
8217
8218    def _parse_grant(self) -> exp.Grant | exp.Command:
8219        start = self._prev
8220
8221        privileges = self._parse_csv(self._parse_grant_privilege)
8222
8223        self._match(TokenType.ON)
8224        kind = self._match_set(self.CREATABLES) and self._prev.text.upper()
8225
8226        # Attempt to parse the securable e.g. MySQL allows names
8227        # such as "foo.*", "*.*" which are not easily parseable yet
8228        securable = self._try_parse(self._parse_table_parts)
8229
8230        if not securable or not self._match_text_seq("TO"):
8231            return self._parse_as_command(start)
8232
8233        principals = self._parse_csv(self._parse_grant_principal)
8234
8235        grant_option = self._match_text_seq("WITH", "GRANT", "OPTION")
8236
8237        if self._curr:
8238            return self._parse_as_command(start)
8239
8240        return self.expression(
8241            exp.Grant,
8242            privileges=privileges,
8243            kind=kind,
8244            securable=securable,
8245            principals=principals,
8246            grant_option=grant_option,
8247        )
8248
8249    def _parse_overlay(self) -> exp.Overlay:
8250        return self.expression(
8251            exp.Overlay,
8252            **{  # type: ignore
8253                "this": self._parse_bitwise(),
8254                "expression": self._match_text_seq("PLACING") and self._parse_bitwise(),
8255                "from": self._match_text_seq("FROM") and self._parse_bitwise(),
8256                "for": self._match_text_seq("FOR") and self._parse_bitwise(),
8257            },
8258        )
8259
8260    def _parse_format_name(self) -> exp.Property:
8261        # Note: Although not specified in the docs, Snowflake does accept a string/identifier
8262        # for FILE_FORMAT = <format_name>
8263        return self.expression(
8264            exp.Property,
8265            this=exp.var("FORMAT_NAME"),
8266            value=self._parse_string() or self._parse_table_parts(),
8267        )
8268
8269    def _parse_max_min_by(self, expr_type: t.Type[exp.AggFunc]) -> exp.AggFunc:
8270        args: t.List[exp.Expression] = []
8271
8272        if self._match(TokenType.DISTINCT):
8273            args.append(self.expression(exp.Distinct, expressions=[self._parse_assignment()]))
8274            self._match(TokenType.COMMA)
8275
8276        args.extend(self._parse_csv(self._parse_assignment))
8277
8278        return self.expression(
8279            expr_type, this=seq_get(args, 0), expression=seq_get(args, 1), count=seq_get(args, 2)
8280        )
8281
8282    def _identifier_expression(
8283        self, token: t.Optional[Token] = None, **kwargs: t.Any
8284    ) -> exp.Identifier:
8285        token = token or self._prev
8286        expression = self.expression(exp.Identifier, this=token.text, **kwargs)
8287        expression.update_positions(token)
8288        return expression